DEV Community: 313 Automations

GEO Checklist: 20 Ways to Get Your Business Cited by ChatGPT and Perplexity

313 Automations — Sun, 25 Oct 2026 13:00:00 +0000

More people now ask ChatGPT, Perplexity, Gemini or Google's AI Overviews for recommendations before they ever click a search result. If those answers never mention your business, you're invisible at the exact moment someone is deciding.

The work of fixing that is often called GEO, generative engine optimization. It builds on SEO rather than replacing it. Nobody can guarantee an AI assistant will cite you, but there are clear signals that make it much easier for them to find, understand and trust your business.

Here's the 20-point checklist we use, grouped into four areas.

1. Entity and identity: make it obvious who you are

AI systems need to be confident that "your business" is one clearly defined thing.

Organization or LocalBusiness schema on the homepage, with name, logo, URL, address, phone, founding date and founder.
sameAs links in that schema pointing to every official profile: LinkedIn company page, Google Business Profile, Crunchbase, Clutch, GitHub and so on.
Identical name, address and phone everywhere: website, Google Business Profile, Bing Places and directories.
A clear About page that says who you are, where you're based, since when, who runs it and what you do.
Author or founder bios with real credentials and links to their profiles.

2. Content: write so answers can be quoted

AI answers are built from passages that are easy to extract and attribute.

Answer-first service pages: open with one or two sentences on what you offer, for whom, where and at what price range.
FAQ sections built from real customer questions, with direct answers.
Specific facts and examples: price ranges, timelines, named case studies. Vague claims are hard to quote.
Honest comparison pages that match how people ask: "X vs Y", "best options for Z".
Visible last-updated dates on key pages.

3. Technical: let the right crawlers in

Don't block the crawlers you want. Review robots.txt for rules affecting search and AI crawlers such as OAI-SearchBot, PerplexityBot, Bingbot and Googlebot. Decide deliberately which ones you allow.
Get indexed in Bing as well as Google. Several AI assistants rely on Bing results for live web answers, so verify your site in Bing Webmaster Tools and submit a sitemap.
Put key content in the HTML, not only in JavaScript, so crawlers can read text, prices and FAQs.
Keep an XML sitemap and use IndexNow or fast re-indexing so changes are picked up quickly.
Consider an llms.txt file. It's an emerging, optional convention with unproven impact, but it's cheap to add.

4. Off-site signals: be mentioned where AI looks

Reviews on trusted platforms like Google, Clutch, G2, Trustpilot or your industry's review sites.
Mentions in articles and best-of lists people trust. Expert quotes in the press help here.
Genuinely helpful answers on Reddit, Quora and communities, since AI assistants often cite these discussions.
Consistent listings on Google Business Profile, Bing Places, Apple Business Connect and quality directories.
Original data, tools or datasets that others reference and link to.

Where to start

If you only do three things this month:

add complete Organization or LocalBusiness schema with sameAs links,
verify your site in Bing Webmaster Tools,
and rewrite the first two sentences of each service page so they answer "what, for whom, where, how much".

Then ask ChatGPT and Perplexity how they describe your business today, write the answers down, and check again in a month.

Free tools to help

The GEO Readiness Checklist turns these 20 checks into an interactive score with your top five priorities.
The Local Business Schema Generator builds LocalBusiness JSON-LD with address, hours, coordinates and sameAs links you can paste into your site.

Both run in your browser and store nothing.

I'm Ghulam Mustafa, founder of 313 Automations, an AI automation, web and SEO/GEO agency in Lahore, Pakistan. Have you checked how AI assistants describe your business? Share what you found in the comments.

We open-sourced 4 repos for AI search and small-business automation

313 Automations — Tue, 20 Oct 2026 10:06:09 +0000

A few weeks ago we shipped free browser tools on Hugging Face. The most common follow-up was "can I use this in my own scripts?" So we put the data, policies and checkers on GitHub, plus two resources we keep sending to clients. Everything is free to use and adapt.

1. ai-crawler-robots-txt

https://github.com/313Automations/ai-crawler-robots-txt

"AI crawlers" are not one thing. Training crawlers (GPTBot, ClaudeBot, CCBot) collect content for models, AI search crawlers (OAI-SearchBot, Claude-SearchBot, PerplexityBot) find pages to cite, and user-triggered fetchers (ChatGPT-User, Perplexity-User) open a page because someone asked. The common mistake is blocking search when you only meant to opt out of training.

The repo has:

data/: 27 crawlers as CSV and JSON, with category, training use, effect of blocking and the operator's docs linked on every row
policies/: three ready-made blocks (allow search and block training, block all AI, allow all)
A checker in JavaScript (Node 18+) and Python 3.8+, no dependencies, that follows RFC 9309 matching: the most specific user-agent group wins, the longest rule wins, Allow wins ties, and * and $ wildcards work

node js/robots-ai-check.mjs https://www.example.com
python python/robots_ai_check.py ./robots.txt /blog/ --json

2. awesome-geo

https://github.com/313Automations/awesome-geo

A curated list for Generative Engine Optimization that sticks to primary sources: Google's AI features guidance, Bing's webmaster guidelines (which now have a GEO section), OpenAI's publisher FAQ, each major AI crawler's own documentation, RFC 9309, llms.txt, Schema.org, IndexNow and the original GEO paper. CC0, and pull requests are welcome.

3. llms-txt-examples

https://github.com/313Automations/llms-txt-examples

Ten fictional businesses (dental clinic, law firm, plumber, restaurant, SaaS, online store, real estate, accountant, marketing agency, nonprofit), one llms.txt each, plus validate_llms_txt.py, a zero-dependency validator for files or live URLs:

python validate_llms_txt.py examples/*.llms.txt
python validate_llms_txt.py https://www.example.com

A note we also make in the README: llms.txt is a proposal. It is cheap to add and can help agents navigate a site, but no major AI search engine has confirmed it as a ranking or citation signal.

4. smb-automation-recipes

https://github.com/313Automations/smb-automation-recipes

30 recipes across sales, marketing, customer support, operations, finance, HR, e-commerce, real estate and clinics. Each one has the trigger and steps, build notes for Zapier, Make and n8n (which built-in tools to reach for: Paths, Routers, Wait nodes, error handlers), a test checklist and the watch-outs that usually bite in production, like duplicate records, SMS consent and keeping a person in the loop for AI output.

Licenses

Code is MIT. Data and text are CC-BY-4.0, and the awesome list is CC0.

All repos: https://github.com/313Automations

Issues and pull requests are welcome. What should we open-source next?

Is your robots.txt blocking AI search? 4 free browser tools we just shipped

313 Automations — Tue, 20 Oct 2026 06:59:20 +0000

We build AI automation and websites for small businesses, and the same questions keep coming up on client calls. So we turned four of them into free, single-file browser tools on Hugging Face. No signup, no backend, nothing stored.

1. AI Crawler robots.txt Checker

The problem: "AI crawlers" are not one thing. Some collect training data (GPTBot, ClaudeBot, CCBot), some build AI search indexes (OAI-SearchBot, Claude-SearchBot, PerplexityBot) and some fetch a page because a user asked (ChatGPT-User, Perplexity-User). Blocking the wrong one can quietly remove a site from AI answers.

The checker parses a pasted robots.txt the way major crawlers do (RFC 9309: most specific user-agent group wins, longest matching rule wins, Allow wins ties, * and $ wildcards) and evaluates 27 crawlers. It then generates a policy for one of three stances: get cited but opt out of training, allow everything, or block AI while keeping Google and Bing.

👉 https://huggingface.co/spaces/313Automations/ai-crawler-robots-checker

The crawler list is also an open dataset with the operator's documentation linked on every row: https://huggingface.co/datasets/313Automations/ai-crawler-user-agents

2. llms.txt Generator

Fill in a name, a one-line summary, key facts and your important pages, and it outputs a valid llms.txt (H1, blockquote, details, H2 link sections, Optional) with checks for missing names and non-https links.

We were honest on the page: Google has said Search doesn't use llms.txt, and no major AI search engine has confirmed it as a ranking or citation signal. It is cheap to add and can help agents navigate, but it is not a substitute for crawlable, clear content.

👉 https://huggingface.co/spaces/313Automations/llms-txt-generator

3. Automate or Hire Calculator

Hours per week, loaded hourly cost, the share automation can handle, build and running cost. It returns two-year cost for both paths, monthly saving, payback and a five-question fit check, then recommends automate, hybrid or hire.

👉 https://huggingface.co/spaces/313Automations/automate-or-hire-calculator

4. AI Missed-Call Responder ROI Calculator

Calls per month, missed rate, share that are new enquiries, conversion, customer value and margin. It shows revenue at risk, what an instant text-back or AI receptionist could win back, and compares profit (not revenue) with the monthly cost.

👉 https://huggingface.co/spaces/313Automations/missed-call-roi-calculator

How they're built

Each tool is one self-contained index.html on a static Hugging Face Space: inline CSS with light and dark themes, vanilla JS, WebApplication JSON-LD, and no external requests. View the source under Files on any Space.

We also published three more datasets: a GEO prompt library (125 buyer questions across 25 industries), 30 SMB workflow automation templates and 45 FAQ/FAQPage examples for service businesses.

Everything lives here: https://huggingface.co/313Automations

What should we build next? Ideas welcome in the comments.

How to Estimate the ROI of an AI Chatbot Before You Build One

313 Automations — Sun, 18 Oct 2026 13:00:00 +0000

"Will an AI chatbot actually pay for itself?" is the first question most business owners ask us, and it's the right one. The honest answer is: it depends on your numbers, and you can estimate them in about ten minutes before spending anything.

Here's the simple model we use.

The five numbers you need

Conversations per month across chat, email, WhatsApp and calls.
The share that are repetitive: order status, prices, opening hours, booking questions, the same FAQs.
The share of those a bot can fully resolve without a person. Be conservative here.
Minutes a person spends per conversation, including looking things up.
Fully loaded staff cost per hour: salary plus overheads, not just the wage.

Then add the costs: a one-time build cost and a monthly running cost (hosting, AI model usage, channel fees such as the WhatsApp Business API, and maintenance).

The calculation

AI-handled conversations = conversations × repetitive share × resolution share
Hours saved per month = AI-handled conversations × minutes per conversation ÷ 60
Gross monthly savings = hours saved × staff cost per hour
Net monthly savings = gross savings − monthly running cost
Payback period = build cost ÷ net monthly savings

That's it. No magic, and every input is something you can measure or estimate yourself.

Where people go wrong

Overestimating the resolution rate

A chatbot that "handles" a conversation but leaves the customer confused hasn't saved anything. It's created a second conversation. Use a conservative resolution share for planning, then measure the real number in the pilot.

Forgetting running costs

Model usage scales with conversations. Channels like WhatsApp can add per-conversation fees. Someone also needs to update the knowledge base when prices or policies change. Put all of that in the monthly cost.

Ignoring the value that isn't time

Faster replies can mean more leads converted and fewer abandoned carts. Out-of-hours answers can capture enquiries you'd otherwise lose. These are real but harder to estimate, so treat them as upside rather than baking them into the payback.

Assuming headcount cuts

In most small businesses the time saved goes into faster responses, sales follow-up and better service, not redundancies. That's usually where the real return is.

How to get real numbers before you build

Export last month's conversations and tag the top 30 questions. This one exercise tells you:

your actual repetitive share,
which questions the bot must answer well,
and gives the bot its first approved knowledge base.

If the repetitive share is low, a chatbot probably isn't your best first automation. Look at lead follow-up or data entry instead.

Run a small pilot

Scope the bot narrowly (one channel, the top questions), give it a clear handover to a person, and run it for two to four weeks. Compare hours spent and response times against your baseline. If the numbers hold, expand. If they don't, you've learned cheaply.

Try the free calculator

We built a free calculator that does this maths for you in USD, PKR, GBP, EUR, AUD or CAD. Change any input and the hours saved, net savings and payback update instantly. It runs in your browser and stores nothing:

AI Chatbot ROI Calculator on Hugging Face

If you're not sure your business is ready for a chatbot at all, start with the two-minute AI Readiness Checker.

I'm Ghulam Mustafa, founder of 313 Automations, an AI automation and software development agency in Lahore, Pakistan. What's stopping you from trying a chatbot? Let me know in the comments.

Zapier vs Make vs n8n: How to Choose the Right Automation Tool for a Small Business

313 Automations — Sun, 11 Oct 2026 13:00:00 +0000

Zapier, Make and n8n all promise the same thing: connect your apps and let the boring work run itself. They're all good. They're also different enough that picking the wrong one can mean a painful rebuild six months later, or a bill that grows faster than the time you save.

We build automations on all three, plus custom code, for small and mid-sized businesses. Here's how we decide which one fits.

The one-line version

Zapier: the easiest to learn, with very broad app coverage. Best for simple, fast wins owned by non-technical staff.
Make: a visual builder that handles branching, loops and data mapping well, usually at a lower cost per step. A strong middle ground for operations teams.
n8n: developer-friendly, can be self-hosted, with strong code and AI-agent building blocks. Best when you have someone technical and want control.
Custom code: when the logic is unique, the volume is very high or no platform fits.

Pricing models, limits and app lists change often, so always check each vendor's current plans before committing.

Six questions that decide it

1. Who will build and maintain it?

This matters more than any feature list. An automation nobody on your team can fix is a liability.

Non-technical staff → Zapier first, Make second.
A tech-comfortable operations person → Make is often the sweet spot.
A developer or technical team → n8n or custom code.

2. How complex are the workflows?

"When X happens, do Y" → any tool works; Zapier is fastest to set up.
Multi-step flows with filters, branches, iterators and data transformation → Make or n8n.
Heavy custom logic, data processing or AI agents that call several tools → n8n or custom code.

3. How much volume will you run?

All three price differently, but the pattern is consistent: simple per-task pricing is convenient at low volume and can become expensive at high volume. If you expect tens or hundreds of thousands of runs a month, model the cost on each platform before you build, and look at self-hosting or code for the heaviest flows.

4. Where must the data live?

If customer data has to stay on your own servers for privacy or compliance reasons, self-hosted n8n or custom code is usually the practical answer. Cloud platforms are fine for most small businesses, but read the data terms either way.

5. How niche are your apps?

Popular SaaS (Google Workspace, HubSpot, Shopify, Slack) is well covered everywhere. Niche apps with an API are easier when the platform lets you make custom HTTP calls comfortably. Internal databases and legacy systems often push you towards n8n or code.

6. What matters most: speed, cost or control?

Speed to launch → Zapier.
Lowest running cost at scale → Make or self-hosted n8n.
Flexibility and control → n8n or custom code.

You don't have to pick just one

Many of the stacks we maintain combine tools: Zapier or Make for simple app-to-app links that the operations team owns, and n8n or code for the AI agents and high-volume flows that need more control. The important thing is to decide deliberately and document who owns what.

A practical way to start

List every manual copy-paste, export and import your team does in a week.
Mark each one with volume, complexity and who would own the automation.
Pick the tool that fits most of the list, not the shiniest feature.
Automate one workflow, add an error alert, and run it for two weeks before adding the next.

Try the free picker

We turned these six questions into a free tool that ranks Zapier, Make, n8n and custom code for your situation and explains why. It runs in your browser and stores nothing:

Automation Tool Picker on Hugging Face

Not sure what to automate first? The AI Use Case Finder lists practical ideas by industry and team.

I'm Ghulam Mustafa, founder of 313 Automations, an AI automation and software development agency in Lahore, Pakistan. Which tool are you using, and what made you choose it? Tell me in the comments.

Is Your Business Ready for AI Automation? A 5-Point Checklist Before You Build

313 Automations — Sat, 03 Oct 2026 20:09:16 +0000

Most small businesses don't fail at AI because the models are weak. They fail because they automate the wrong thing first, or they automate a process that only exists in someone's head.

We've been building automation and web projects for small and mid-sized businesses since 2018, and the same pattern keeps showing up: the projects that pay off start small, start with a boring repetitive task, and start with the groundwork already done.

Here's the checklist we use before recommending any AI build. You can run through it in about two minutes.

1. Processes: is the work repeatable and written down?

AI copies whatever process you give it, including the messy parts. If three people answer the same customer question three different ways, a chatbot will happily learn all three.

Ask yourself:

Is the task done the same way every time?
Could a new hire follow written steps to do it?

If not: write the steps down first. A one-page checklist is enough. This is the single cheapest thing you can do to make an AI project succeed.

2. Data: does the AI have something reliable to work from?

A support chatbot is only as good as its sources. If your FAQs live in old emails, WhatsApp chats and one person's memory, the bot has nothing trustworthy to answer from.

Ask yourself:

Are your FAQs, policies, prices and product details in one up-to-date place?
Is your customer data in a CRM or store that people actually keep current?

If not: collect your top 30 customer questions and their approved answers. That list becomes the bot's first knowledge base, and it shows you how repetitive your enquiries really are.

3. Tools: can your systems talk to each other?

Most useful automations move data between tools: a form to a CRM, an order to a spreadsheet, a lead to a calendar. That's much easier when your tools have APIs or connectors for platforms like Zapier, Make or n8n.

Ask yourself:

Do you run on mainstream tools (Google Workspace, HubSpot, Shopify, WooCommerce)?
Have you automated anything before, even simple email rules?

If not: that's fine, but expect more integration work, and prefer tools with good API support when you next switch.

4. Team: will someone own it and review it?

The most common reason a working automation quietly dies is that nobody owns it. Prices change, the bot gives an outdated answer, and trust disappears.

Ask yourself:

Is there a named person with a few hours a week to own this?
Will someone review AI outputs for the first few weeks?

If not: name an owner before you build. Start with AI drafting and a human approving, then reduce review only where the output has proven reliable.

5. Goals: do you know what success looks like?

"Save time" isn't a goal you can measure. "Reply to every new lead within five minutes" is.

Ask yourself:

What is the one number this should improve?
Do you know today's baseline?

If not: measure it for two weeks before building. Without a baseline you can't tell whether the project worked.

Pick the right first automation

Once the groundwork is in place, choose the task where your team loses the most time each week. In our experience the best first projects usually fall into one of four buckets:

Answering the same customer questions → an AI support assistant grounded in your own documents (a RAG chatbot), with a clear handover to a person
Following up leads → a lead follow-up agent that replies quickly, asks a few qualifying questions and books calls
Copying data between tools → no-code integrations with error alerts
Scheduling and reminders → automated confirmations and reminders to cut no-shows

Keep the first pilot narrow, run it for two to four weeks, and compare against your baseline.

Try the free checklist

We turned this checklist into a free, two-minute tool. It scores your business across the five areas above, shows your weakest two, and suggests the best first automation for your biggest time sink. It runs in your browser and doesn't store anything:

AI Readiness Checker on Hugging Face

If you want ideas for your industry, the AI Use Case Finder lists practical use cases for e-commerce, real estate, healthcare, SaaS, fintech and professional services, each with typical tools and a first step.

The short version

Write the process down.
Put your information in one reliable place.
Use tools that can connect.
Name an owner and keep a human in the loop.
Pick one measurable goal and record the baseline.

Do those five things and almost any sensible AI pilot has a fair chance. Skip them and even the best model will struggle.

I'm Ghulam Mustafa, founder of 313 Automations, an AI automation and software development agency in Lahore, Pakistan. Questions about where AI fits in your business? Ask in the comments and I'll answer.