DEV Community: Ievgenii Gryshkun

Magento 2 AEO Guide: Make Your Store Visible in ChatGPT, Gemini and Perplexity (2026)

Ievgenii Gryshkun — Wed, 29 Apr 2026 20:51:10 +0000

TL;DR — 2-minute version

Default Magento 2 stores are largely invisible to AI assistants (~25% AEO score)
The biggest blockers: robots.txt blocks AI bots, no llms.txt, Product schema missing offers.availability
You can audit your store in seconds: bin/magento angeo:aeo:audit
Most stores reach 80%+ AEO in ~90 minutes with basic fixes
ChatGPT Shopping requires a separate merchant registration at chatgpt.com/merchants

You've optimized your Magento store for Google. But when a shopper asks ChatGPT "what's a good cast iron pan under $80?" — your store isn't in the answer.

That's not an SEO problem. It's an AEO problem. And the fixes are completely different.

AI assistants are becoming a product discovery layer — not just a search tool. When a user asks for a recommendation, they've already formed intent. If your store isn't accessible to these systems, you're invisible at exactly that moment. That's a different problem from ranking fifth in Google — it's not being in the conversation at all.

This guide covers all 9 AEO signals for Magento 2, what each one means, and how to fix every gap — with exact CLI commands. The tooling referenced is open-source and on Packagist, but the signal framework applies to any Magento implementation.

Key terms

AEO — AI Engine Optimization
The practice of configuring a website so AI assistants — ChatGPT, Gemini, Perplexity — can discover, read, and recommend it. Different from SEO: targets AI-generated answers, not Google Search rankings. Relevant signals: robots.txt access, structured data, llms.txt, ACP product feeds.

Q: Is AEO the same as GEO (Generative Engine Optimization)?
Largely yes — both terms describe optimization for AI-generated results. AEO is used primarily in the eCommerce context; GEO is more common in content publishing. The underlying signals are the same.

OAI-SearchBot
OpenAI's live search crawler. Used when ChatGPT answers real-time queries. Respects robots.txt. Must be explicitly allowed — default Magento robots.txt blocks it via wildcard.

Q: Is OAI-SearchBot the same as GPTBot?
No. OAI-SearchBot is for live query answering. GPTBot is for training data collection. Blocking GPTBot has no effect on ChatGPT Shopping visibility.

ACP — Agentic Commerce Protocol
OpenAI's open standard for structured merchant product data. Defines a .jsonl.gz feed format submitted at chatgpt.com/merchants — required for ChatGPT product results.

llms.txt
A plain-text file at yourstore.com/llms.txt — a structured map of your catalog for AI systems. Spec at llmstxt.org. Similar to sitemap.xml, but optimized for AI consumption.

offers.availability
A required field in Product JSON-LD schema. Tells AI systems whether a product is in stock. Value: https://schema.org/InStock or https://schema.org/OutOfStock. Missing on most default Magento installations — its absence fails ChatGPT Shopping conformance checks.

AEO vs SEO: what's different

SEO gets your store ranked in Google's link list. AEO gets your store cited in AI-generated answers.

Google reads keywords, backlinks, and page authority. AI systems read:

robots.txt — can AI crawlers access your store at all?
llms.txt — is there a structured map of your catalog?
Product JSON-LD schema — can AI read prices, availability, and product details?
ACP product feed — have you registered with OpenAI's merchant program?

Most of these don't exist on a default Magento 2 install — not because Magento is broken, but because it was built before AI search existed.

Q: Does a store with good SEO automatically have good AEO?
No. Most signals don't overlap. A store can rank on page one of Google and still score near zero on AI product feed and llms.txt — because those signals didn't exist when the store was built.

Signal	Helps SEO	Helps AEO
robots.txt AI bot access	—	✅
llms.txt / llms.jsonl	—	✅
Product JSON-LD schema	✅	✅
FAQPage schema	✅	✅
ACP product feed	—	✅
sitemap.xml	✅	✅
Open Graph tags	✅	✅
Canonical tags	✅	✅

The 9 AEO signals for Magento 2

Run the free audit to see your current score:

composer require angeo/module-aeo-audit
bin/magento setup:upgrade
bin/magento angeo:aeo:audit

About the weights below: each signal has a weight (0.5–1.0) reflecting its relative impact on overall AEO score. Critical signals (robots.txt, schema, feed) are weight 1.0; supporting signals are lower. Failing a 1.0-weighted signal alone can drop you below the 50% threshold.

Signal #1 — robots.txt: AI Bot Access (weight 1.0)

The most common reason for complete AI invisibility. Magento's default robots.txt uses a wildcard rule (User-agent: *) combined with restrictive Disallow directives. Since AI crawlers like OAI-SearchBot, GPTBot, and PerplexityBot aren't explicitly listed, they inherit those restrictions and get blocked unintentionally.

Per OpenAI's crawler documentation, stores that block OAI-SearchBot may significantly limit or prevent visibility in ChatGPT search results.

This is the most frequently failing signal across all store sizes — and the cheapest to fix.

Check your file at yourstore.com/robots.txt. You need these entries, placed before any wildcard block:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Google-Extended
Allow: /

User-agent: ClaudeBot
Allow: /

Quick fix — two options:

Option A: edit robots.txt manually in your Magento admin (Stores → Configuration → Design → Search Engine Robots) or directly in the file. Make sure your edits actually persist (see the nginx warning below).

Option B: install a module that injects and validates the rules:

composer require angeo/module-robots-txt-aeo
bin/magento setup:upgrade && bin/magento cache:flush
bin/magento angeo:robots:validate

⚠️ Adobe Commerce Cloud: robots.txt is served via Fastly VCL. After any change — purge Fastly cache and verify the live file, not the admin config.

Signal #2 — llms.txt: AI Content Map (weight 1.0)

llms.txt is a plain-text file at yourstore.com/llms.txt — a structured map of your catalog: categories, key products, CMS pages, currency, language. Defined at llmstxt.org.

Perplexity's background indexer actively reads it. For ChatGPT Shopping, it supplements the ACP product feed.

You can hand-write llms.txt for a small catalog (the format is just markdown), but for stores with hundreds+ of products it gets stale fast. A common automation approach:

composer require angeo/module-llms-txt
bin/magento setup:upgrade
bin/magento angeo:llms:generate

This also generates llms.jsonl — the machine-readable sibling. Enable cron for automatic regeneration when your catalog changes.

→ Deep dive: How to generate llms.txt for Magento 2 in 5 minutes

Signal #3 — Product JSON-LD Schema (weight 1.0)

Magento's default Luma theme outputs basic Product schema but almost always omits offers.availability — a key field for ChatGPT Shopping conformance. Submissions without it commonly fail validation.

Q: Does Hyvä theme fix the schema problem automatically?
Partially. Hyvä has better schema defaults than Luma but still requires explicit configuration for offers.availability and variant-level data. Neither platform handles this out of the box — a dedicated module or theme override is required.

Quick check on any product page:

curl -s https://yourstore.com/sample-product | grep -o '"availability":"[^"]*"'

You can fix this either via a custom theme override or with a module. Example using a module:

composer require angeo/module-rich-data
bin/magento setup:upgrade && bin/magento cache:flush

This adds: offers.availability (real-time stock), aggregateRating, BreadcrumbList. It also covers Signal #5 — FAQPage schema on CMS pages.

The 4 most common schema failures I see (in order of frequency):

Missing offers.availability — 90%+ of default Magento stores. Without it, ChatGPT Shopping conformance fails outright.
Variant-level data flattened — configurable products lose variesBy data when serialized. ChatGPT shows "out of stock" even when 6 of 8 variants are available.
Hyvä strips microdata — performance optimization removes JSON-LD entirely. Needs a layout XML override to re-inject.
Wrong availability values — using "InStock" (string) instead of "https://schema.org/InStock" (URI). Validators accept it, AI engines silently drop it.

If your audit shows Signal #3 failing, run this diagnostic:

# Replace with any product URL from your store
curl -s https://yourstore.com/sample-product | \
  grep -A 20 'application/ld+json' | \
  grep -E 'availability|priceCurrency|sku'

If availability is missing or doesn't start with https://schema.org/, that's your problem.

Example — Magento 2 store, home goods, ~2,400 SKUs

⚠️ Before — default install:

AEO score: 23%
✗ robots.txt — OAI-SearchBot blocked
✗ llms.txt — not found
✗ Product schema — no availability
✗ AI product feed — not registered

✅ After — 90 minutes later:

AEO score: 84%
✓ robots.txt — 7 AI bots allowed
✓ llms.txt — 2,400 products mapped
✓ Product schema — availability live
◷ Feed — submitted, pending approval

💡 At this point — after signals #1, #2, and #3 — most stores jump from low-twenties into the 50–60% range. Run the audit again to confirm your progress before continuing:
bin/magento angeo:aeo:audit

Signal #4 — AI Product Feed / ChatGPT Shopping (weight 1.0)

ChatGPT Shopping requires a registered ACP product feed. Without it, products are unlikely to appear in ChatGPT Shopping results regardless of other signals. The feed is .jsonl.gz, submitted to a private SFTP endpoint provided by OpenAI after approval, refreshed every 15 minutes.

Building an ACP feed from scratch is doable but non-trivial — you need to handle JSONL streaming, gzip compression, SFTP auth, the offers.availability mapping, and 15-minute refresh cron. Most teams either build it in-house or use a module:

composer require angeo/module-openai-product-feed \
  angeo/module-openai-product-feed-api
bin/magento setup:upgrade
bin/magento angeo:aeo:feed:generate
bin/magento angeo:aeo:feed:validate

Apply at chatgpt.com/merchants. Currently US-only and waitlisted. After approval, set up 15-minute cron — stale availability data is the #1 reason for post-approval suppression.

→ Deep dive: How to Register Your Magento 2 Store for ChatGPT Shopping

Does this actually move the needle?

Fair question — and the honest answer is "yes, but unevenly across platforms."

Perplexity shows the fastest impact. Background re-indexing typically picks up llms.txt and schema changes within days to a few weeks.
ChatGPT is gated by merchant approval. Technical fixes alone won't get you into Shopping results — you also need the feed registered and approved. Once approved, indexing is usually 48–72 hours.
Gemini mostly piggybacks on Google Merchant Center. If you're already there, fixes propagate quickly. If not, getting into Merchant Center is the bigger lift than AEO itself.

AEO is not instant. But with these signals failing, visibility across all three engines is close to zero — fixing them is the floor, not the ceiling.

Signals #5–#9 — Supporting Signals

FAQPage schema (weight 0.5)
Increases citation probability for answer-style AI queries. Injected automatically alongside Product schema on CMS pages. Verify:

curl -s yourstore.com | grep FAQPage

sitemap.xml (weight 0.8)
AI crawlers use sitemap.xml to discover your full catalog. Enable in Magento admin: Stores → Config → Catalog → XML Sitemap. Submit to Google Search Console — also surfaces to Google-Extended (Gemini).

Open Graph tags (weight 0.7)
og:title, og:description, og:image — used as content fallback when structured schema is absent. Most Magento themes include these by default.

Canonical tags (weight 0.6)
Prevents AI systems from indexing Magento's multiple URL variants. Enable: Stores → Config → Catalog → SEO → Use Canonical Link Meta Tag For Products/Categories.

Note: Magento has no canonical option for the homepage — this is expected.

llms.jsonl (weight 0.75)
Machine-readable catalog at yourstore.com/llms.jsonl. One JSON object per product per line. Used by AI pipelines for vector indexing. Generated automatically by angeo/module-llms-txt.

ChatGPT vs Perplexity vs Gemini: how each AI discovers products

ChatGPT — hybrid approach: trained knowledge + live search via OAI-SearchBot + ACP merchant product feed. Requires merchant registration. Feed is the authoritative source for product data.

Perplexity — real-time answer engine. Always cites sources with links. PerplexityBot actively reads llms.txt — one of the few AI platforms where this has documented crawl value. No separate product feed required.

Gemini — operates on Google's Shopping Graph (50B+ products, updated 2B+ times per hour). Any product in Google Merchant Center is available to Gemini. Google-Extended is the separate crawler for AI Overviews — allowing it does not affect regular Google Search rankings.

Cheat sheet — what each engine actually reads:

Source	ChatGPT	Perplexity	Gemini
ACP product feed	✅ primary	—	—
OAI-SearchBot crawl	✅ secondary	—	—
llms.txt	⚠️ untested	✅ documented	⚠️ untested
Product JSON-LD	✅	✅	✅ via Shopping Graph
Google Merchant Center	—	—	✅ primary
Open Graph tags	✅ fallback	✅ fallback	✅ fallback
robots.txt	`OAI-SearchBot`, `GPTBot`, `ChatGPT-User`	`PerplexityBot`	`Google-Extended`

The pattern: ChatGPT prefers structured submissions (the feed), Perplexity prefers crawl-discoverable structure (llms.txt + schema), Gemini piggybacks on Google's existing pipelines (Shopping Graph + Merchant Center). Cover all three and you're visible across the major engines.

Score interpretation

Score	Label	Typical situation
0–25%	Needs Improvement	Default Magento install. AI crawlers blocked.
26–50%	Needs Improvement	Some fixes applied. Schema or feed still missing.
51–75%	Moderate	Core signals in place. Feed not yet registered.
76–90%	Good	Strong foundation. Minor schema gaps.
91–100%	Excellent	Full compliance. Feed registered and live.

Free open-source modules — one per signal

Each AEO signal has a corresponding open-source Magento 2 module. MIT licensed. Install via Composer. Runs on your server — no external API calls, no data sent outside your infrastructure.

Module	Signal
`angeo/module-aeo-audit`	Diagnostic — checks all 9 signals, score trend dashboard
`angeo/module-robots-txt-aeo`	Signal #1 — injects AI bot rules into robots.txt
`angeo/module-llms-txt`	Signals #2, #2b — generates llms.txt and llms.jsonl
`angeo/module-rich-data`	Signals #3, #5 — Product JSON-LD + FAQPage schema
`angeo/module-openai-product-feed`	Signal #4 — ACP product feed for ChatGPT Shopping

90-minute implementation checklist

Baseline audit: bin/magento angeo:aeo:audit — note your starting score
Fix robots.txt: composer require angeo/module-robots-txt-aeo → validate with angeo:robots:validate
Generate llms.txt: composer require angeo/module-llms-txt → angeo:llms:generate
Fix Product schema: composer require angeo/module-rich-data → verify offers.availability in page source
Enable sitemap: Stores → Config → Catalog → XML Sitemap → Enable → submit to Search Console
Generate product feed: composer require angeo/module-openai-product-feed angeo/module-openai-product-feed-api
Apply for ChatGPT Shopping: configure seller name + policy URLs → apply at chatgpt.com/merchants
Final audit: bin/magento angeo:aeo:audit → target 80%+

Adobe Commerce Cloud: additional steps

Adobe Commerce Cloud adds one complication — Fastly CDN. robots.txt is stored in pub/media/ and served via a Fastly VCL snippet. After any robots.txt change:

Save in Magento admin
Purge Fastly cache
Verify the live file at yourstore.com/robots.txt — not the admin config

Fastly also caches product pages. If availability changes but Fastly serves a stale page, OAI-SearchBot reads incorrect stock data — leading to product suppression after ChatGPT Shopping approval. Use the 15-minute cron for the ACP product feed as the authoritative source for availability.

Adobe Commerce vs Magento Open Source — AEO-relevant differences:

Open Source: robots.txt is editable directly, cron is yours, no Fastly cache layer to invalidate. AEO setup is straightforward.
Adobe Commerce Cloud: Fastly is mandatory, robots.txt is served via VCL, cron has timing constraints. Every AEO change needs an extra "purge Fastly" step.
B2B Commerce on Adobe Commerce: customer-group pricing breaks ACP feed assumptions (the protocol expects single price per SKU). You'll need a separate "public" feed with B2C pricing for ChatGPT Shopping. Don't expose B2B pricing to AI engines.

If you're on Adobe Commerce Cloud, expect implementation to take 2-3x longer than vanilla Magento — most of the extra time is fighting Fastly cache TTLs, not AEO logic itself.

Frequently asked questions

Does AEO replace SEO for Magento stores?

No. They target different channels. SEO gets you ranked in Google Search. AEO gets you cited in AI-generated answers. A store with strong SEO can still score very low on AEO by default — because the signals are completely different.

Is ChatGPT Shopping free for merchants?

Product discovery is currently organic — no cost to appear in results. No transaction fees on purchases that start in ChatGPT as of April 2026. Onboarding is US-only and waitlisted.

How long before results are visible after fixing AEO?

Technical fixes take effect as AI crawlers re-index your pages: 1–3 weeks for Perplexity, 2–6 weeks for Google-Extended, 48–72 hours after ChatGPT Shopping approval for the feed layer.

What is GPTBot vs OAI-SearchBot?

GPTBot is OpenAI's training crawler. Blocking it opts your content out of GPT model training but has no effect on ChatGPT Shopping visibility. OAI-SearchBot is the live search crawler — this one must be allowed.

Does this work with Hyvä theme?

Yes. All Angeo modules are compatible with Magento Open Source and Adobe Commerce on both Luma and Hyvä themes. PHP 8.2+ required.

Start here

Run the audit:

bin/magento angeo:aeo:audit

If your score is below 50%, your store is likely invisible to AI assistants today. Fixing the first 3 signals (robots.txt, llms.txt, Product schema) usually gets you most of the way there.

Install CLI Audit — MIT-licensed, runs on your server, no telemetry
Free Web Self-Assessment — no install required, runs against your live site

llms.txt for Magento 2: What It Is, Why It Matters, and How to Generate It in 5 Minutes

Ievgenii Gryshkun — Fri, 24 Apr 2026 04:48:48 +0000

Google has robots.txt. Google has sitemap.xml. Now AI assistants have llms.txt.

If you run a Magento 2 store and haven't heard of llms.txt yet — this post is the 5-minute version of everything you need to know, plus the exact commands to generate it today.

The problem: AI assistants don't crawl like Google does

When Google indexes your store, its crawler visits thousands of pages and builds a picture of your catalog over weeks. It's slow, but it works.

ChatGPT, Claude, and Perplexity operate differently. When a user asks "what's a good blender under $150?", the AI doesn't crawl your store in real time. It works from what it already knows — or from a structured snapshot it can parse in milliseconds.

If you haven't given it that snapshot, it either ignores your store or gets the details wrong: outdated prices, discontinued products, categories that no longer exist.

llms.txt is that snapshot.

What is llms.txt?

llms.txt is a plain-text file hosted at the root of your domain — yourstore.com/llms.txt — that gives AI systems a structured, human-readable summary of your store.

Think of it as a README.md for your entire eCommerce site. It tells AI assistants:

What your store sells
What categories exist
Key product URLs and descriptions
CMS pages (shipping, returns, about)
Store metadata: name, currency, language

A minimal example looks like this:

# Store: Volta Equipment

## STORE
Name: Volta Equipment
URL: https://voltaequipment.com/
Currency: USD

### CATEGORIES ###
Category ID: 4
Name: Fitness Equipment
URL: https://voltaequipment.com/fitness
Description: Commercial and home gym equipment

### PRODUCTS ###
SKU: VE-KB32
Name: 32kg Kettlebell — Cast Iron
Price: 8900
URL: https://voltaequipment.com/32kg-kettlebell
Short Description: Competition-grade cast iron kettlebell, powder-coated finish

## CMS PAGES
TYPE: PAGE
TITLE: Shipping Policy
URL: https://voltaequipment.com/shipping

Simple. No JSON schema. No special syntax. Just structured plain text that both AI systems and humans can read instantly.

Does llms.txt actually work?

Fair question — and the honest answer is: it depends on the AI.

Some AI systems (Perplexity, some Claude configurations) actively read llms.txt when crawling. Others, like GPTBot, primarily rely on the page content itself and Product schema JSON-LD. OpenAI has not yet published an official spec for llms.txt consumption.

So why generate it at all?

Three reasons:

1. It costs nothing to add. One Composer command and it's live. There's no downside.

2. It's a trust signal. When a Magento developer or agency reviews your AEO setup, a present llms.txt signals you're thinking about AI visibility — and it's one of the 8 signals checked by the AEO audit module.

3. The standard is evolving fast. OpenAI, Anthropic, and Google are all moving toward structured merchant data formats. Stores with llms.txt already in place will require zero additional work when these systems formalize their specs. Stores without it will have to catch up.

llms.txt vs llms.jsonl — what's the difference?

The angeo/module-llms-txt module generates both files. Here's when each matters:

File	Format	Best for
`llms.txt`	Plain text	AI assistants reading store context, Perplexity, Claude
`llms.jsonl`	JSON Lines	Structured product data, future ACP integrations, developers

llms.jsonl is the machine-readable sibling — one JSON object per product, per line, compressed for fast parsing. If you're planning to integrate with OpenAI's ACP feed later, llms.jsonl is the stepping stone.

For now, generate both. They serve different consumers.

Installation: 3 commands

composer require angeo/module-llms-txt
bin/magento setup:upgrade
bin/magento cache:flush

That's it. The module is MIT licensed, free on Packagist, and runs entirely on your server.

Generate the files

Option A — CLI (recommended)

bin/magento angeo:llms:generate

This generates both llms.txt and llms.jsonl at your store root and outputs the file paths.

Option B — Magento Admin

Go to Stores → Configuration → General → Angeo → LLMS and trigger manual generation from the backend.

Option C — Cron (for automatic updates)

The module registers a Magento cron job. As long as your server cron is running, the files stay current:

bin/magento cron:run

For stores with frequently changing inventory — new products, price updates, seasonal categories — cron is the right default.

Verify it's live

After generation, open yourstore.com/llms.txt in a browser. You should see your store name, category list, and product entries.

Then run the AEO audit to confirm it passes signal #2:

bin/magento angeo:aeo:audit

Expected output:

✓ PASS  robots.txt — OAI-SearchBot and GPTBot allowed
✓ PASS  llms.txt — store content map present
...

If llms.txt shows FAIL, check that the file exists at pub/llms.txt and that your web server is serving it correctly (not blocked by nginx or a CDN rule).

Multi-store support

If you're running multiple store views or storefronts, the module generates separate llms.txt files per store — each with the correct currency, language, and URL structure. No additional configuration needed.

What to do after llms.txt

llms.txt is AEO signal #2. To complete your AI visibility setup, the remaining signals are:

✅ robots.txt — Allow OAI-SearchBot, GPTBot, ClaudeBot, and 4 others
✅ llms.txt — structured store map ← you are here
→ Product Schema — JSON-LD with offers.availability on every product page
→ AI Product Feed — ACP-compliant .jsonl.gz for ChatGPT Shopping registration
→ FAQPage Schema — structured Q&A for policy pages

Run bin/magento angeo:aeo:audit to see your current score across all 8 signals.

Summary

llms.txt gives AI assistants a structured snapshot of your store
It's served at yourstore.com/llms.txt — same pattern as robots.txt
The angeo/module-llms-txt module generates it automatically, with cron support
It takes 3 commands and under 5 minutes to set up
It's one of 8 AEO signals that determine your visibility in ChatGPT, Claude, and Perplexity

Free modules used in this post:

angeo/module-llms-txt — generates llms.txt and llms.jsonl
angeo/module-aeo-audit — checks all 8 AEO signals via CLI

Run a free AEO audit on your store →

How to Register Your Magento 2 Store for ChatGPT Shopping (ACP Feed + Step-by-Step)

Ievgenii Gryshkun — Mon, 20 Apr 2026 17:20:06 +0000

ChatGPT now processes over 50 million shopping queries per day. Appearing in those results is not automatic — it requires a deliberate application process, a spec-compliant product feed, and passing OpenAI's conformance checks.

This guide walks through the entire process: from the pre-check that determines if your store is ready, to the merchant application, to the ongoing cron setup that keeps your feed current.

Shopify and Etsy merchants already have native integrations. This guide is for Magento 2 / Adobe Commerce stores.

🔗 Originally published on angeo.dev

Before You Apply: The Pre-Check

OpenAI's conformance checks verify technical requirements before granting production access. Stores that fail the pre-check go into a review queue and receive vague rejection emails. Do the pre-check yourself first.

bin/magento angeo:aeo:audit

Expected output before you start:

✓ PASS  robots.txt — OAI-SearchBot and GPTBot allowed
✓ PASS  llms.txt — store content map present
✓ PASS  Product Schema — JSON-LD with offers.availability
✗ FAIL  AI Product Feed — no feed found

Signal #5 (AI Product Feed) will fail until Step 1 below. Signals #1, #2, and #3 must all PASS before applying at chatgpt.com/merchants.

If robots.txt is failing — most default Magento 2 installs have a User-agent: * wildcard that silently blocks OAI-SearchBot. Fix that first. It's a hard gate that invalidates everything else.

Step 1 — Install the product feed modules

composer require angeo/module-openai-product-feed \
  angeo/module-openai-product-feed-api

bin/magento setup:upgrade
bin/magento cache:flush

Both modules are MIT licensed, free on Packagist. No license keys, no SaaS — runs entirely on your server.

Step 2 — Configure seller information

Go to Stores → Configuration → Angeo → Product Feed API and fill in:

Seller name
Target country — ISO 3166 two-letter code (e.g. US)
Policy URLs — privacy policy, returns, shipping, terms of service

These populate seller.links on every product record. They are merchant credibility signals in OpenAI's scoring. Missing policy URLs is a common reason for suppressed products after initial approval.

Step 3 — Generate and validate the feed

# Generate the feed file
bin/magento angeo:aeo:feed:generate

# Validate structure and required fields
bin/magento angeo:aeo:feed:validate

Confirm in the output:

Each product has a variants array
Prices are in minor units (see the price format section below)
availability.status is present on every product
seller.links has at least two policy URLs

Step 4 — Verify promotions output

bin/magento angeo:aeo:feed:validate --type=promotions

Active Magento cart price rules should appear automatically as ACP promotions with active_period, benefits, and status: "active". If promotions are missing here, they won't appear in ChatGPT results either.

Step 5 — Apply at chatgpt.com/merchants

Submit your store URL and business details. OpenAI tests schema compliance, HTTP response codes, and feed integrity.

Current status: Onboarding is US-only and available to approved partners — expect a waitlist. Application review typically takes 1–2 weeks.

Step 6 — Receive SFTP endpoint and push the feed

After approval, OpenAI provides a private SFTP endpoint for your store. Push your .jsonl.gz feed file to this endpoint.

⚠️ Do not host the feed publicly on your website. The feed goes to the private SFTP endpoint OpenAI provides — not a public REST endpoint on your store. Submit a sample first; after it passes validation, push the full catalog.

Step 7 — Set up 15-minute cron

# crontab — every 15 minutes
*/15 * * * * /usr/bin/php /var/www/html/bin/magento angeo:aeo:feed:generate --push

OpenAI accepts feed refreshes every 15 minutes. Full feed re-submission is required each time — there is currently no incremental update support.

⚠️ Most common post-approval failure: Stale availability data — out-of-stock products showing as available — is the #1 reason products get suppressed after initial approval. The 15-minute cron prevents this.

Feed Format: What OpenAI Actually Accepts

Format	Extension	Notes
JSON Lines	`.jsonl.gz`	Recommended. One product object per line. Handles nested variants cleanly.
CSV	`.csv.gz`	Works for flat catalogs. Variant structures need to be flattened.

Note: TSV and XML were in OpenAI's original spec announcement but have since been removed. Always check the official feed spec for the current list.

The Price Format Trap

This is the single most common implementation error across all ACP integrations.

Price must be sent in ISO 4217 minor units as an integer:

# ❌ Wrong — fails validation
"price": 14.99
"price": "14.99"

# ✅ Correct — integer minor units
"price": 1499    // €14.99
"price": 14900   // $149.00

The module handles this automatically: (int) round($price * 100). But if you are building a custom integration or mapping from a Google Shopping feed — check this field first. The validation error message for a wrong price format is non-obvious.

What OpenAI's Conformance Check Validates

Check	What it verifies	Handled by
Feed schema	Required fields, correct data types, valid URIs	ProductMapper
Price format	Integer minor units, ISO 4217 currency code	ProductMapper
Availability flags	`enable_search` and `availability.status` present	ProductMapper
Promotion schema	`benefits` array with type, dates, `status: "active"`	PromotionMapper
Seller links	At least 2 policy URLs present	Admin config → seller.links
Product IDs	Unique, stable, no duplicates across the feed	ProductMapper

No Fees for Discovery

Product discovery results in ChatGPT are currently organic and unsponsored. There is no cost to submit a product feed or appear in shopping results.

The 4% transaction fee announced with Instant Checkout applied only to completed in-chat purchases — and OpenAI is moving away from that model toward merchant-owned checkout. As of April 2026, there are no fees on purchases that start in ChatGPT.

After approval, the feed needs to be indexed — typically 48–72 hours. Products with complete Product JSON-LD schema, high availability accuracy, rich descriptions, and active promotions appear first.

Checklist Before Submitting

[ ] bin/magento angeo:aeo:audit — robots.txt, llms.txt, Product schema all PASS
[ ] Feed generates without errors and produces valid .jsonl.gz
[ ] Each product has variants, correct prices in minor units, availability.status
[ ] Seller info complete: name, country (ISO 3166), at least 2 policy URLs
[ ] Active promotions appear with status: "active"
[ ] 15-minute cron configured and tested

All four Angeo modules are free, MIT licensed, and available on Packagist:

angeo/module-aeo-audit — AEO CLI audit, 8 signals
angeo/module-llms-txt — auto-generates llms.txt
angeo/module-openai-product-feed — ACP feed generator
angeo/module-openai-product-feed-api — REST API, 6 endpoints

Free AEO Self-Assessment →

How to Fix robots.txt for ChatGPT and Gemini in Magento 2 (2026 Update)

Ievgenii Gryshkun — Thu, 16 Apr 2026 20:41:06 +0000

Your sitemap is configured. Your Core Web Vitals score is green. Your product catalog is perfectly structured. And yet when a user asks ChatGPT for products you sell, your store doesn't appear.

Most of the time, the reason is a single file: robots.txt.

Specifically — a robots.txt written for Google in 2019 and never updated for the ten AI crawlers that now determine your visibility in ChatGPT, Gemini, Claude, and Perplexity.

🔗 Originally published at angeo.dev/magento-2-robots-txt-chatgpt-gemini-ai-bots

Why robots.txt Is AEO Signal #1

The angeo/module-aeo-audit checks robots.txt first and marks it Critical because it is a gate. Every other AEO signal — llms.txt, Product schema, AI product feed — is irrelevant if the AI crawler cannot enter your store.

OpenAI states this without ambiguity:

"Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers."

Not "may not appear." Will not appear. If OAI-SearchBot is blocked — by an explicit Disallow or caught in a wildcard rule — your store is excluded from ChatGPT search answers regardless of everything else.

The Three Types of AI Bots — Why the Difference Matters

Before listing every bot, understand what each one actually does. Conflating them causes the most common robots.txt misconfiguration.

Type	What it does	Examples	Recommendation
Search & indexing	Builds the live index used when users ask AI questions. Cites sources, links back to your store.	`OAI-SearchBot`, `Claude-SearchBot`, `PerplexityBot`, `Google-Extended`	✅ Always allow
User-initiated	Fetches your page when a user asks AI to visit a specific URL. May cite your product page directly.	`ChatGPT-User`, `Claude-User`, `Perplexity-User`	✅ Always allow
Training crawlers	Collects content for model training. No attribution, no traffic back.	`GPTBot`, `ClaudeBot`, `Applebot-Extended`	⚠️ Your choice

The most common mistake: blocking GPTBot (training) thinking it removes you from ChatGPT search results. It does not. GPTBot and OAI-SearchBot are entirely separate bots. Blocking training crawlers has zero effect on AI search visibility — but blocking search crawlers makes you invisible immediately.

Every AI Bot That Matters in 2026

Bot	Platform	Type	Impact if blocked
`OAI-SearchBot`	ChatGPT	Search index	Invisible in all ChatGPT search answers. Most critical.
`GPTBot`	ChatGPT	Training	Excluded from future GPT training. No effect on current search.
`ChatGPT-User`	ChatGPT	User-initiated	ChatGPT can't fetch your pages for users.
`Claude-SearchBot`	Claude	Search index	Invisible in Claude's real-time web search answers.
`ClaudeBot`	Claude	Training	Excluded from future Claude training data.
`Claude-User`	Claude	User-initiated	Claude can't fetch your pages for users.
`PerplexityBot`	Perplexity	Search index	Invisible in Perplexity answers and recommendations.
`Perplexity-User`	Perplexity	User-initiated	Perplexity can't fetch your pages for users.
`Google-Extended`	Gemini	Search + training	Not cited in Gemini AI Overviews or Google Shopping AI.
`Applebot-Extended`	Apple Intelligence	Training	Excluded from Apple Intelligence training data.
`anthropic-ai`	Anthropic	Deprecated	Legacy name for ClaudeBot. Keep rules for backwards compatibility.

Anthropic expanded to three bots in early 2026. Sites that only reference ClaudeBot are now missing Claude-SearchBot (live search) and Claude-User (user-initiated fetching). If your robots.txt hasn't been updated since 2024, this almost certainly applies to your store.

The Default Magento robots.txt Problem

Magento's default robots.txt starts with a wildcard:

User-agent: *
Disallow: /index.php/
Disallow: /*?
Disallow: /checkout/
Disallow: /app/
...

This wildcard establishes a baseline that every bot inherits. If your deployment script, hosting provider, or a staging migration has added Disallow: / anywhere — AI bots are caught in it silently, with no error logged anywhere.

Check this right now. Open https://yourstore.com/robots.txt and look for Disallow: / on its own line. If it exists without an explicit Allow: / for each AI bot listed above it — every one of those bots is blocked.

# Quick check — if this returns output, you have a problem
curl -s https://yourstore.com/robots.txt | grep -n "^Disallow: /$"

Where Magento Serves robots.txt — Two Scenarios

Before editing, identify which method your store uses. Editing the wrong one has no effect.

Scenario A: Magento Admin (most common)

# Check if Magento manages robots.txt
bin/magento config:show design/search_engine_robots/default_robots

If it returns a value, edit via: Content → Design → Configuration → [Store view] → Edit → Search Engine Robots → Edit custom instruction of robots.txt

Scenario B: Static file in pub/

# Check if a static file exists and is being served
ls -la /var/www/html/pub/robots.txt
curl -I https://yourstore.com/robots.txt
# If no X-Magento headers appear, the file is served statically

Edit pub/robots.txt directly, or remove it to let Magento Admin take over.

Multi-store: Each store view can have its own robots.txt. If you run multiple stores on different domains, configure each store view separately in Admin → Design → Configuration.

The Complete robots.txt for Magento 2

Paste this as your full configuration. AI bot entries must appear before the wildcard User-agent: * block.

# ============================================================
# AI SEARCH & INDEXING BOTS — Always allow
# Blocking these makes your store invisible in AI search answers.
# ============================================================
User-agent: OAI-SearchBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Google-Extended
Allow: /

# ============================================================
# USER-INITIATED FETCHERS — Always allow
# ============================================================
User-agent: ChatGPT-User
Allow: /

User-agent: Claude-User
Allow: /

User-agent: Perplexity-User
Allow: /

# ============================================================
# TRAINING CRAWLERS — your choice
# Blocking does NOT affect search visibility.
# Change Allow to Disallow to opt out of training data collection.
# ============================================================
User-agent: GPTBot
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: anthropic-ai
Allow: /

User-agent: Applebot-Extended
Allow: /

# ============================================================
# TRADITIONAL SEARCH ENGINES
# ============================================================
User-agent: Googlebot
Allow: /

User-agent: Bingbot
Allow: /

# ============================================================
# ALL OTHER BOTS — standard Magento rules
# AI bots above are explicitly allowed before this wildcard.
# ============================================================
User-agent: *
Allow: /

# Magento paths — block from all crawlers
Disallow: /admin/
Disallow: /adminhtml/
Disallow: /api/
Disallow: /rest/
Disallow: /graphql
Disallow: /cron.php
Disallow: /var/
Disallow: /lib/
Disallow: /dev/
Disallow: /index.php/
Disallow: /*?SID=
Disallow: /*?___store=
Disallow: /checkout/
Disallow: /customer/
Disallow: /wishlist/
Disallow: /review/

# ============================================================
# SITEMAPS
# ============================================================
Sitemap: https://yourstore.com/sitemap.xml
Sitemap: https://yourstore.com/llms.txt

⚠️ Order is not optional. robots.txt uses first-match semantics per crawler. If User-agent: * with Disallow: / appears before the AI bot entries, those AI bots are permanently blocked — the rules below are never reached.

Four Mistakes That Block AI Bots

❌ Mistake 1 — Disallow: / left on from staging

Many Magento stores use Disallow: / on staging. This is frequently copied to production and never removed.

Common sources:

Magento Admin: Stores → Configuration → General → Design → Search Engine Robots
Manually edited pub/robots.txt copied from staging
CI/CD pipelines that sync the full staging filesystem to production
Managed hosts (Hypernode, Nexcess, Cloudways) applying restrictive defaults on new environments

❌ Mistake 2 — Wildcard block placed before AI bot rules

# ❌ WRONG — wildcard fires first, AI bot rules below it are ignored
User-agent: *
Disallow: /

User-agent: OAI-SearchBot
Allow: /  # ← never reached, bot already matched the wildcard above

# ✅ CORRECT — explicit AI rules appear before the wildcard
User-agent: OAI-SearchBot
Allow: /

User-agent: *
Disallow: /checkout/

❌ Mistake 3 — Using outdated Anthropic bot names

# Deprecated — no longer reflect Anthropic's bot infrastructure
User-agent: Claude-Web     # retired 2024
User-agent: Anthropic-AI   # retired 2024

# Current names (2026)
User-agent: ClaudeBot        # training
User-agent: Claude-SearchBot # live search index ← missing from most configs
User-agent: Claude-User      # user-initiated fetching ← missing from most configs

❌ Mistake 4 — robots.txt served from cache or static file

Some hosting setups bypass Magento when serving robots.txt:

Nginx static rule serves pub/robots.txt before Magento handles the request
CDN caching — Cloudflare or Fastly caches the old file with long TTLs
Varnish returns cached response without hitting the application

# Diagnosis — if no X-Magento headers, file is served statically
curl -I https://yourstore.com/robots.txt

Fix for Nginx — ensure this location block is present:

location = /robots.txt {
    try_files $uri $uri/ /index.php$is_args$args;
}

Fix for Cloudflare — purge cache for /robots.txt via dashboard, or add a Cache Rule to bypass caching for that path.

Verify the Fix

# Check all critical AI search bots have Allow: /
curl -s https://yourstore.com/robots.txt | grep -A1 -E \
  "OAI-SearchBot|Claude-SearchBot|PerplexityBot|Google-Extended"

# Check server logs to confirm bots are actually crawling
grep -Ei "OAI-SearchBot|Claude-SearchBot|PerplexityBot" \
  /var/log/nginx/access.log | tail -20

# Via AEO audit module — checks all bots + validates rule order
composer require angeo/module-aeo-audit
bin/magento setup:upgrade
bin/magento angeo:aeo:audit

# ✓ PASS  robots.txt — AI Bot Access
#         OAI-SearchBot ✓  Claude-SearchBot ✓  ChatGPT-User ✓
#         ClaudeBot ✓  Claude-User ✓
#         PerplexityBot ✓  Google-Extended ✓

# Multi-store — run per store view
bin/magento angeo:aeo:audit --store=de
bin/magento angeo:aeo:audit --store=fr

After robots.txt — What's Next

Fixing robots.txt is the access layer. Once AI bots can reach your store, the signals that determine whether you actually appear in AI answers are:

llms.txt — machine-readable index of your catalog that AI assistants parse in seconds → angeo/module-llms-txt
Product JSON-LD schema — structured markup that tells AI exactly what your products cost and where to buy them → checked by angeo/module-aeo-audit
AI product feed — structured feed required for ChatGPT Shopping eligibility → angeo/module-openai-product-feed

# See all 8 AEO signals and your complete score
bin/magento angeo:aeo:audit

FAQ

Q: Does blocking GPTBot affect ChatGPT recommendations?

No — but blocking OAI-SearchBot does. These are separate bots. GPTBot collects training data; OAI-SearchBot builds the index ChatGPT uses for search answers. Blocking GPTBot has no effect on whether your store appears in ChatGPT results. Blocking OAI-SearchBot removes you from ChatGPT search answers entirely.

Q: What is OAI-SearchBot and why does it matter more than GPTBot?

OAI-SearchBot is OpenAI's crawler that powers ChatGPT real-time search results and product recommendations. It determines whether your store appears in ChatGPT answers today. GPTBot collects training data for future model versions — its effects are long-term and indirect. Both should be allowed, but OAI-SearchBot is the one that determines your current search visibility.

Q: How do I check which AI bots are blocked in Magento 2?

curl https://yourstore.com/robots.txt | grep -i \
  'searchbot\|gptbot\|claude\|perplexity\|google-extended'

If nothing appears, bots may be blocked by a wildcard Disallow: / rule. The fastest automated check:

composer require angeo/module-aeo-audit
bin/magento setup:upgrade
bin/magento angeo:aeo:audit

It checks all bots, validates rule order, and reports exact status for each one.

Q: Should I allow training bots like GPTBot and ClaudeBot?

For most ecommerce stores, yes. Allowing training crawlers means your product descriptions and category content contribute to how AI models understand and describe products in your category — which indirectly improves recommendation quality over time. If you have legal or privacy reasons to opt out, block only the training bots (GPTBot, ClaudeBot, Applebot-Extended) — your search visibility is not affected.

Check all AI bots and 7 other AEO signals in one command — free, MIT licensed:

📦 Install AEO Audit Module
🌐 Free Web Self-Assessment

Originally published at angeo.dev

How to Check If Your Magento Store Is Visible to ChatGPT — Free AEO Audit Module

Ievgenii Gryshkun — Tue, 14 Apr 2026 20:35:57 +0000

Part 2 of the AEO for Magento 2 series. Part 1: What is llms.ljson and why eCommerce needs it.

Most Magento developers spend hours manually checking robots.txt, schema markup, and sitemap configurations. Then they wonder why their store still doesn't appear in ChatGPT or Gemini recommendations.

There is now a faster way.
🔗 This article was originally published on https://angeo.dev/magento-aeo-audit-module-chatgpt-visibility/:

The problem: no standard AEO checklist for Magento

SEO has Lighthouse. SEO has Search Console. SEO has audit tools built into every CI pipeline.

AEO — AI Engine Optimization — has almost nothing.

Magento developers who want to know whether their store is AI-visible have to check manually:

Is GPTBot allowed in robots.txt?
Does the store have a llms.txt file?
Is Product JSON-LD schema present on PDP pages?
Is there an AI-readable product feed for ChatGPT Shopping?
Does the homepage have FAQPage schema for answer-box eligibility?

This takes time, requires domain knowledge, and produces no consistent score to track over time.

Introducing: angeo/module-aeo-audit

angeo/module-aeo-audit is a free, open-source Magento 2 CLI module that runs a complete AEO audit with one command.

composer require angeo/module-aeo-audit
bin/magento setup:upgrade
bin/magento angeo:aeo:audit

The module checks 8 signal categories, scores your store from 0–100, and tells you exactly what to fix — with specific commands.

What it checks

Signal	Category	What it looks for
robots.txt — AI Bot Access	Critical	GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended
llms.txt — AI Content Map	Critical	Spec-compliant llms.txt at /llms.txt
Product JSON-LD Schema	Important	@type Product with offers.price and offers.availability
FAQPage Schema	Important	FAQPage schema on homepage
AI Product Feed	Important	Feed file at expected paths
sitemap.xml	Standard	sitemap_index.xml or sitemap.xml
Open Graph	Standard	og:title, og:description, og:image
Canonical Tags	Standard	rel=canonical on all pages

What the output looks like

╔══════════════════════════════════════════════════════════╗
║  Angeo AEO Audit — Store: default                        ║
╚══════════════════════════════════════════════════════════╝

✓ PASS  robots.txt — All 7 AI bots are permitted.
✗ FAIL  llms.txt — /llms.txt not found (404).
✓ PASS  sitemap.xml — found (1,243 URLs).
✓ PASS  Product Schema — Product JSON-LD found.
⚠ WARN  FAQPage Schema — No FAQPage schema on homepage.
✗ FAIL  AI Product Feed — No AI-readable product feed found.
✓ PASS  Open Graph — All required OG tags found.
✓ PASS  Canonical Tags — Canonical tag found.

AEO Score: [████████████░░░░░░░░] 50% — Needs Improvement
✓ Pass: 5   ⚠ Warn: 1   ✗ Fail: 2

Critical fixes needed:
→ composer require angeo/module-llms-txt
→ bin/magento angeo:llms:generate
→ composer require angeo/module-openai-product-feed

Every failed check includes a specific command. Not "fix your schema" — but exactly what to run.

Three output formats

# Default — readable table in terminal
bin/magento angeo:aeo:audit

# JSON — for dashboards or automated processing
bin/magento angeo:aeo:audit --format=json --output=/tmp/aeo-report.json

# Markdown — for Notion, docs, or sharing with clients
bin/magento angeo:aeo:audit --format=markdown

# Specific store only
bin/magento angeo:aeo:audit --store=en_gb

CI pipeline integration

The --fail-on flag exits with code 1 if the score drops below a threshold:

# Fail the build if AEO score drops below 80
bin/magento angeo:aeo:audit --fail-on=80

Add this to your GitHub Actions or GitLab CI pipeline — if a deploy breaks AI visibility, the build fails. Same way phpstan or phpcs works for code quality.

# .github/workflows/aeo.yml
- name: AEO Audit
  run: bin/magento angeo:aeo:audit --fail-on=80

How it fits the Angeo suite

module-aeo-audit is the diagnostic layer. It tells you what's missing. The other free modules fix those gaps:

# Fix signal 2 — llms.txt
composer require angeo/module-llms-txt
bin/magento angeo:llms:generate

# Fix signal 5 — AI product feed
composer require angeo/module-openai-product-feed

# Re-run audit to verify
bin/magento angeo:aeo:audit

Installation

Requirements: PHP 8.2+, Magento 2.4+, Magento Open Source or Adobe Commerce Cloud.

composer require angeo/module-aeo-audit
bin/magento setup:upgrade
bin/magento cache:flush
bin/magento angeo:aeo:audit

No admin configuration. No database changes. One CLI command registered.

Why this exists

ChatGPT referral traffic converts at 4–5× the rate of standard organic search. Most Magento stores have at least 2–3 critical AEO issues they don't know about — a robots.txt that blocks OAI-SearchBot, a product page with no JSON-LD schema, no llms.txt while competitors already have one.

The audit takes 30 seconds. The fixes usually take less than an hour.

Next in series: The Complete AEO Checklist for Magento 2 — 8 Signals, 90 Minutes — the step-by-step fix guide for every signal the audit checks.

Module on Packagist: packagist.org/packages/angeo/module-aeo-audit

What is llms.jsonl and why eCommerce needs it

Ievgenii Gryshkun — Sun, 12 Apr 2026 08:10:18 +0000

AI assistants don't read your website the way customers do. They need structured, machine-readable data — and that's exactly what llms.jsonl provides.

If you already know about llms.txt, think of llms.jsonl as its structured, data-rich sibling — purpose-built for eCommerce product catalogs.

🔗 This article was originally published on https://angeo.dev/what-is-llms-jsonl-and-why-ecommerce-needs-it/:

What Is llms.jsonl?

llms.jsonl stands for LLM-optimized Line-delimited JSON. It is a file where each line contains a complete, self-contained JSON object representing a single entity — a product, a category, or a CMS page.

Unlike a regular JSON array (which wraps everything in brackets and requires loading the whole file at once), JSONL is:

Streamable — AI agents can read it line by line
Scalable — stores with 50,000 SKUs don't need one giant file
Parseable — each line is valid JSON independently

A Single Product Line in llms.jsonl

{"type":"product","sku":"WB-004","name":"Alpine Hiking Jacket","price":189.99,"currency":"USD","url":"https://angeo.test/alpine-jacket","category":"Outerwear","short_description":"Waterproof 3-layer shell for alpine conditions","in_stock":true,"attributes":{"color":"Navy","size_options":["S","M","L","XL"],"material":"Gore-Tex"}}

Every product — one line. Every line — complete context for an AI to understand and recommend that product.

llms.txt vs llms.jsonl — What Is the Difference?

Feature	llms.txt	llms.jsonl
Format	Markdown (human-readable)	Line-delimited JSON (machine-readable)
Best for	Store overview, pages, categories	Full product catalog data
AI use case	Context & navigation	Product recommendations & queries
Scalability	Up to ~500 items	Unlimited (streamed line by line)
Attributes	Basic (name, URL, price)	Full (variants, stock, specs)

Both files work together. llms.txt tells AI what your store is. llms.jsonl tells AI what your store sells — in full detail.

Why Does eCommerce Specifically Need llms.jsonl?

1. AI Agents Need Structured Product Data

When a user asks ChatGPT or Claude "recommend a waterproof jacket under $200," the AI needs structured, queryable product data — not HTML product pages. llms.jsonl provides exactly that: a clean feed of every product with its attributes, price, stock status, and URL.

2. HTML Product Pages Are Noise for AI

A typical Magento 2 product page contains navigation menus, cookie banners, review widgets, upsell carousels, and footer links. The actual product data is buried inside this noise. AI context windows are limited, and parsing messy HTML wastes them. llms.jsonl gives AI only signal, zero noise.

3. Product Catalogs Are Too Large for llms.txt

A store with 5,000 SKUs cannot fit them all meaningfully into a single llms.txt file. llms.jsonl handles any catalog size because AI systems can stream and process it line by line — one product at a time.

4. Variants, Stock, and Pricing Change Daily

Unlike static blog content, product data changes constantly. A configurable product might have 12 size/color combinations. Stock goes in and out. Prices change with promotions. llms.jsonl is generated fresh by cron and always reflects the current state of your catalog.

What Does a Full llms.jsonl File Look Like?

Each line is an independent JSON object. Here is a multi-entity example:

{"type":"store","code":"angeo_en","name":"EN","url":"https:\/\/angeo.test\/","currency":"USD","locale":en}
{"type":"category","store":"angeo_nl","id":"4","name":"Sale","parent_id":"2","url":"https:\/\/angeo.test\/sale.html","description":"","embedding_text":"Sale "}
{"type":"product","store":"angeo_en","id":"4","sku":"product_sku","title":"product name","price":"123.00","currency":"USD","short_description":"","description":"test descriprtion","url":"https:\/\/angeo.test\/product-name.html","embedding_text":"product name  test descriprtion"}

An AI agent processing this file can instantly answer:

"What waterproof jackets do you have under $300?" → WB-001
"Do you have merino base layers?" → WB-002
"Is the 3-in-1 jacket in stock?" → No (in_stock: false)

How to Generate llms.jsonl for Magento 2

The open-source module angeo/module-llms-txt generates both llms.txt and llms.jsonl automatically for your Magento 2 store.

Install in 3 commands:

composer require angeo/module-llms-txt
bin/magento setup:upgrade
bin/magento cache:flush

After installation, navigate to:

Stores → Configuration → General → Angeo → LLMS

Generate manually or let cron handle scheduled updates. Both files are created at your store root:

https://yourstore.com/llms.txt
https://yourstore.com/llms.jsonl

The module supports multi-store and multi-language Magento 2 setups, includes unit tests for both file types, and is MIT licensed.

Source: Packagist

The Bigger Picture: AI Commerce Infrastructure

We are entering a phase where AI agents don't just recommend products — they initiate purchase flows on behalf of users. For an AI agent to recommend your products, it must be able to:

Understand what you sell
Filter by price, attributes, stock status
Navigate to a product URL and initiate checkout

Steps 1 and 2 are solved by llms.jsonl. Step 3 is the next frontier — but without steps 1 and 2, step 3 is impossible.

Stores that generate structured AI feeds today will be the ones AI agents recommend tomorrow.

Summary

llms.jsonl is a line-delimited JSON file that exposes your full product catalog to AI systems
It complements llms.txt — together they give AI both context and structured data
For Magento 2, the open-source angeo/module-llms-txt generates both files automatically
Implementation takes under 5 minutes and requires no custom development

The AI commerce era is not coming — it is already here. Structured product feeds are the foundation that makes your store visible to the systems that are replacing traditional search.

Have questions about llms.jsonl for Magento 2? Drop a comment below or check the module on Packagist.

DEV Community: Ievgenii Gryshkun

Magento 2 AEO Guide: Make Your Store Visible in ChatGPT, Gemini and Perplexity (2026)

TL;DR — 2-minute version

Key terms

AEO vs SEO: what's different

The 9 AEO signals for Magento 2

Signal #1 — robots.txt: AI Bot Access (weight 1.0)

Signal #2 — llms.txt: AI Content Map (weight 1.0)

Signal #3 — Product JSON-LD Schema (weight 1.0)

Example — Magento 2 store, home goods, ~2,400 SKUs

Signal #4 — AI Product Feed / ChatGPT Shopping (weight 1.0)

Does this actually move the needle?

Signals #5–#9 — Supporting Signals

ChatGPT vs Perplexity vs Gemini: how each AI discovers products

Score interpretation

Free open-source modules — one per signal

90-minute implementation checklist

Adobe Commerce Cloud: additional steps

Frequently asked questions

Does AEO replace SEO for Magento stores?

Is ChatGPT Shopping free for merchants?

How long before results are visible after fixing AEO?

What is GPTBot vs OAI-SearchBot?

Does this work with Hyvä theme?

Start here

Related posts in this series

llms.txt for Magento 2: What It Is, Why It Matters, and How to Generate It in 5 Minutes

The problem: AI assistants don't crawl like Google does

What is llms.txt?

Does llms.txt actually work?

llms.txt vs llms.jsonl — what's the difference?

Installation: 3 commands

Generate the files

Verify it's live

Multi-store support

What to do after llms.txt

Summary

How to Register Your Magento 2 Store for ChatGPT Shopping (ACP Feed + Step-by-Step)

Before You Apply: The Pre-Check

Step 1 — Install the product feed modules

Step 2 — Configure seller information

Step 3 — Generate and validate the feed

Step 4 — Verify promotions output

Step 5 — Apply at chatgpt.com/merchants

Step 6 — Receive SFTP endpoint and push the feed

Step 7 — Set up 15-minute cron

Feed Format: What OpenAI Actually Accepts

The Price Format Trap

What OpenAI's Conformance Check Validates

No Fees for Discovery

Checklist Before Submitting

How to Fix robots.txt for ChatGPT and Gemini in Magento 2 (2026 Update)

Why robots.txt Is AEO Signal #1

The Three Types of AI Bots — Why the Difference Matters

Every AI Bot That Matters in 2026

The Default Magento robots.txt Problem

Where Magento Serves robots.txt — Two Scenarios

The Complete robots.txt for Magento 2

Four Mistakes That Block AI Bots

❌ Mistake 1 — Disallow: / left on from staging

❌ Mistake 2 — Wildcard block placed before AI bot rules

❌ Mistake 3 — Using outdated Anthropic bot names

❌ Mistake 4 — robots.txt served from cache or static file

Verify the Fix

After robots.txt — What's Next

FAQ

How to Check If Your Magento Store Is Visible to ChatGPT — Free AEO Audit Module

The problem: no standard AEO checklist for Magento

Introducing: angeo/module-aeo-audit

What it checks

What the output looks like

Three output formats

CI pipeline integration

How it fits the Angeo suite

Installation

Why this exists

What is llms.jsonl and why eCommerce needs it

What Is llms.jsonl?

A Single Product Line in llms.jsonl

llms.txt vs llms.jsonl — What Is the Difference?