DEV Community: Rida Moideen

Evaluate AI Visibility Tools Like a Product Manager: A Scoring Script for Enterprise SEO Teams

Rida Moideen — Sun, 08 Nov 2026 15:59:29 +0000

Search any list of top ai visibility tools for enterprise seo teams and you'll find feature grids. Feature grids are how vendors want to be compared. They're rarely how your team should compare them.
This post takes a positioning-first approach: decide what job the tool is being hired to do, for whom, instead of what alternative, and then score tools against that. It includes a small, honest scoring script. The scenario is an illustrative private hospital group in Doha whose SEO team has been asked to report AI visibility to leadership.
Step 1: Define the job, not the features
Start with Kiyado's guide to measuring GEO. It sets out seven KPIs: brand mention rate, citation rate, competitor visibility, brand accuracy, citation quality, AI-referred traffic and conversions. A tool is only useful insofar as it helps you measure these reliably.
For a hospital group, two of those matter more than usual:
Brand accuracy. Healthcare information must be correct. If an AI assistant describes a department or service that a hospital doesn't offer, that's a patient-safety and trust issue, not just a marketing one. Any content changes made in response must be accurate and compliant with the healthcare advertising rules that apply in Qatar.
Citation quality. Patients and referring doctors should see the hospital described from authoritative, current sources.
So the job is: "Help us know, every month, whether AI describes our hospitals accurately for the questions patients ask, and who appears instead."
Step 2: Name the real alternative
If the team doesn't buy a tool, what does it do? Usually a fixed list of prompts, run by hand, recorded in a spreadsheet, as the guide describes: create a prompt set, record the baseline, repeat the test, connect to business data, find the gaps.
That manual baseline is your competitor. A tool has to beat it on the dimensions that matter to you, or it isn't worth the licence.
Step 3: Write your criteria as config
Keep evaluation criteria in version control, weighted by what matters for your job.

criteria.yaml : weights must sum to 100

job: "Monthly AI visibility and accuracy reporting for hospital services"
criteria:
fixed_prompt_sets: {weight: 15, question: "Can we define, version and freeze our own prompts?"}
mention_vs_citation: {weight: 15, question: "Are mentions and citations reported separately?"}
accuracy_workflow: {weight: 20, question: "Can clinicians or comms mark answers inaccurate, with notes?"}
source_visibility: {weight: 15, question: "Can we see and rate the cited sources?"}
competitor_tracking: {weight: 10, question: "Can we track named competitors per prompt?"}
raw_export: {weight: 10, question: "Can we export raw rows to join with traffic and enquiries?"}
collection_compliance: {weight: 10, question: "Is answer collection within each AI provider's terms?"}
arabic_and_english: {weight: 5, question: "Can prompts run in both Arabic and English?"}
Notice what's not there: a single "AI visibility score". The guide is explicit that GEO success can't be reduced to a single number.
Step 4: Score honestly
import yaml

def score(criteria_path, ratings):
"""ratings: {tool: {criterion: 0..3}}, where 0 = no, 1 = partly, 2 = yes, 3 = better than our manual baseline"""
crit = yaml.safe_load(open(criteria_path))["criteria"]
assert sum(c["weight"] for c in crit.values()) == 100, "weights must sum to 100"
results = {}
for tool, r in ratings.items():
missing = [k for k in crit if k not in r]
if missing:
raise ValueError(f"{tool}: unrated criteria {missing}")
results[tool] = round(sum(crit[k]["weight"] * r[k] / 3 for k in crit), 1)
return dict(sorted(results.items(), key=lambda x: -x[1]))

ratings = {
"manual_baseline": {k: 2 for k in ["fixed_prompt_sets", "mention_vs_citation", "accuracy_workflow",
"source_visibility", "competitor_tracking", "raw_export",
"collection_compliance", "arabic_and_english"]},
# "vendor_a": {...}, "vendor_b": {...} fill in after trials
}
print(score("criteria.yaml", ratings))
Rate the manual baseline too. If a vendor doesn't clearly beat it on your highest-weighted criteria, keep the spreadsheet a little longer.
Step 5: Trial with your own prompts
Never evaluate a tool on the vendor's demo prompts. Load your fixed prompt set and run both the tool and your manual process for the same month. Compare:
Do they agree on mentions and citations?
Does the tool surface accuracy issues your team found by hand?
Can you export the rows and join them to enquiries and appointment requests?
The guide's step four, connecting visibility to business data, is where many tools fall short. Test it explicitly.
Where audit tools fit
The guide mentions Geoptie and SEOmator, which assess things like citation readiness, schema, E-E-A-T and AI crawler access. These are diagnostic tools: they suggest why you're not cited. Monitoring tools tell you whether you are. Score them on separate criteria; they're hired for different jobs. Verify current capabilities before you rely on any of them.
Step 6: Segment the buyers inside the hospital group
"The SEO team" is rarely the only buyer of this data. In a hospital group, at least four groups care about different parts of it:
Leadership wants a monthly trend: are we appearing more for the questions patients ask, and are appointment requests following?
Clinical and medical affairs want accuracy alerts: is any AI answer describing a service, specialty or doctor incorrectly?
Communications want source quality: which outdated listings or articles are AI engines relying on?
The SEO team wants prompt-level detail: which prompts, which competitors, which cited pages.
Add a short criterion for each group to criteria.yaml, or score tools separately per audience. A tool that delights the SEO team but gives clinicians no way to flag errors isn't the right tool for a hospital, whatever its feature list says.
Step 7: Decide what "winning" means before the trial
Write down, before any trial starts, what result would make you buy. For example: "The tool must match our manual mention and citation results on at least most prompts, surface every accuracy issue our reviewers found, and export rows we can join to appointment requests." If you write this afterwards, it's too easy to talk yourself into the tool with the nicest charts.
Honest limits
AI answers vary between runs and by location. Any tool's numbers are estimates.
Collection methods matter. Prefer vendors with official API access or transparent, terms-compliant collection.
No tool replaces a human review of accuracy in healthcare.
Summary
Define the job. Name the real alternative. Weight what matters for your team. Trial on your own prompts. The best tool isn't the top of someone else's list. It's the one that does your job better than the spreadsheet does.
Kiyado works remotely with healthcare and enterprise teams in Doha to set up prompt sets and measurement before any tool is chosen.
See how Kiyado helps healthcare organisations with AI visibility →

Your Security Stack Is Hiding Your Product From AI Search (and How to Fix It Without Weakening Anything)

Rida Moideen — Thu, 29 Oct 2026 05:32:48 +0000

Here's an irony I see often at cybersecurity SaaS companies.
Your product is excellent at blocking bad bots. Your own website uses it, or something like it. And somewhere in the rules, it's also blocking the crawlers that AI search platforms use to read your pages.
Meanwhile, marketing is asking why ChatGPT recommends three competitors for "endpoint detection for mid-sized Canadian companies" and never mentions you.
Kiyado's article on how LLMs choose which sources to cite lists six signals. The sixth is the simplest and the most often broken: access. Engines can't cite pages they can't reach.
If you're positioning against bigger vendors and looking for the best geo agency for cybersecurity saas, this is the first thing they should check. You can check it yourself today.
Start with positioning
Before touching configs, be clear on what you want AI engines to cite you for. In security, generic claims ("enterprise-grade protection") get ignored. Specific positioning gets repeated:
"Example Shield is managed detection and response for Canadian mid-market companies with 200–2,000 employees, with a SOC based in Toronto and data residency in Canada."
That sentence is your target. Every technical fix below exists so engines can read and repeat it.
Step 1: Check robots.txt
The article recommends allowing the crawlers AI platforms use, such as OpenAI's OAI-SearchBot.
curl -s https://www.example-security.ca/robots.txt
Look for blanket disallows or crawler-specific blocks:
User-agent: *
Disallow: /
or
User-agent: OAI-SearchBot
Disallow: /
A reasonable pattern is to allow search-oriented crawlers on public marketing and docs pages while keeping private paths blocked:
User-agent: OAI-SearchBot
Allow: /
Disallow: /app/
Disallow: /portal/

User-agent: *
Disallow: /app/
Disallow: /portal/
Check each AI platform's current documentation for crawler names and what each one does. Some crawlers are for search and citation; others serve different purposes. Decide per crawler, deliberately.
Step 2: Check the WAF and bot management
robots.txt is only a request. Your WAF or CDN decides who actually gets through. Security companies often have aggressive bot scoring that challenges or blocks anything non-human.
Test from outside your network:
curl -s -o /dev/null -w "%{http_code}\n" \
-A "Mozilla/5.0 (compatible; OAI-SearchBot/1.0; +https://openai.com/searchbot)" \
https://www.example-security.ca/product/mdr/
A 403, 429 or a challenge page means the crawler is likely being blocked. Note that user-agent strings can be spoofed, so serious bot management verifies crawlers by published IP ranges or reverse DNS. Use the provider's verification method rather than allowing a user-agent string blindly.
Then add a narrowly scoped rule: verified AI search crawlers, allowed on public marketing and docs paths only. Nothing else changes.
Step 3: Get facts out of PDFs
Security vendors love PDFs: datasheets, compliance summaries, analyst excerpts, solution briefs. The article recommends checking whether key facts sit on HTML pages rather than only inside PDFs.

Find pages that link to PDFs but have little text of their own

curl -s https://www.example-security.ca/compliance/ | grep -o 'href="[^"]*.pdf"'
For each important PDF, put the key facts on the page itself: certifications (SOC 2 Type II, ISO 27001), data residency, supported platforms, response-time commitments. Link the PDF as supporting evidence.
Step 4: Fix consistency
The article's fifth signal: when sources conflict, engines hedge or skip the claim.
Security vendors are prone to this. The website says "24/7 SOC," a partner listing says "business-hours monitoring" (from an old plan), an analyst profile lists last year's product name. Build a facts file and render key pages from it:
product: "Example Shield MDR"
soc_location: "Toronto, Canada"
coverage: "24/7/365"
data_residency: "Canada"
certifications: ["SOC 2 Type II", "ISO 27001"]
target_customers: "Canadian companies with 200–2,000 employees"
Then update marketplace listings, partner pages and review profiles to match.
Step 5: Add verifiable evidence
The article reports that the research that named GEO found the biggest gains came from adding statistics, citations and credible quotes. In security, that means certifications with dates, measured response times (if you publish them), customer counts and named frameworks you align with. Avoid unverifiable superlatives.
Step 6: Measure
Run a fixed set of prompts across ChatGPT, Gemini and Perplexity:

"Best MDR providers for mid-sized companies in Canada"
"MDR services with Canadian data residency"
"What is Example Shield?" Record named, cited and accurate, plus which sources were cited. After fixing access, it can take time for engines to recrawl. Measure monthly. The other signals Access gets you read. The rest gets you chosen: Rankings help less than you'd expect: only about 11% of AI assistant citations matched the top 10 Google or Bing results. Target the sub-questions. Brand mentions outweigh backlinks: 0.664 vs 0.218 correlation across 75,000 brands. Analyst notes, security press and conference talks count. YouTube mentions correlated most strongly of all. Freshness: cited content was 25.7% fresher on average. Keep threat and product pages current. Common objections from security teams "Allowing crawlers increases our attack surface." Scope the rule to verified crawlers, public paths and read-only requests. Your app, portal and API stay behind the same controls as before. "We don't want our content used for training." That's a legitimate choice. Many platforms publish separate crawlers for search and for training, so you can allow one and block the other. Read each provider's documentation and decide per crawler. "Marketing can just publish the facts somewhere else." They can, but your own site is the source you control. If it's unreachable, engines rely on third parties that may be out of date. The point You don't have to weaken your security posture to be visible. You need narrowly scoped, verified access for search crawlers on public pages, and facts that engines can read and trust. Kiyado offers a free audit, including an access check: SEO and GEO for Toronto businesses →

Building an Enterprise AI Visibility Monitoring Pipeline: Prompt Sets, Logging Schema and Dashboards

Rida Moideen — Wed, 28 Oct 2026 03:48:15 +0000

Building an Enterprise AI Visibility Monitoring Pipeline: Prompt Sets, Logging Schema and Dashboards
Enterprise marketing teams in Singapore and across APAC are asking a new question: "How do AI assistants describe us in each market, and is it getting better?"
Answering it properly needs a monitoring pipeline, not ad-hoc screenshots. This post walks through a practical design. For the concepts behind it, read Kiyado's explainer on GEO first. It's also a sensible baseline before you evaluate the best generative engine optimization platforms on the market.
Design goals

Repeatable: the same prompts, engines and schedule every cycle.
Multi-market: prompts per country and language.
Auditable: raw answers stored, not just scores.
Actionable: every issue maps to a GEO layer (content, identity or corroboration). Step 1: Version-controlled prompt sets Store prompts in Git so changes are reviewed and dated: # prompts/sg.yml market: SG language: en prompts:
- id: sg-cat-01 text: "Who are the leading {category} providers in Singapore?" type: category
- id: sg-brand-01 text: "What is {brand} known for in Southeast Asia?" type: brand
- id: sg-comp-01 text: "{brand} vs {competitor} for enterprises in Singapore" type: comparison Keep the prompt set stable. If you change it, version it, or your trend lines become meaningless. Step 2: A logging schema CREATE TABLE ai_visibility_runs ( run_date DATE NOT NULL, market VARCHAR(8) NOT NULL, engine VARCHAR(32) NOT NULL, prompt_id VARCHAR(32) NOT NULL, prompt_text TEXT NOT NULL, answer_text TEXT NOT NULL, named BOOLEAN, cited BOOLEAN, accurate BOOLEAN, issue_layer VARCHAR(16), -- content | identity | corroboration notes TEXT, PRIMARY KEY (run_date, market, engine, prompt_id) ); Storing answer_text is essential. When accuracy drops, you need the exact wording to trace the source. Step 3: Collection Collect answers via official APIs where available, or manually for engines without one. Respect each provider's terms of service and rate limits. Note that API outputs can differ from consumer apps, so document which surface you're measuring. def score(answer: str, brand_aliases: list[str], domains: list[str]) -> dict: a = answer.lower() return { "named": any(alias.lower() in a for alias in brand_aliases), "cited": any(d.lower() in a for d in domains), } Use brand aliases (legal name, short name, regional names) and all owned domains. Enterprises often have several. Step 4: Accuracy review Automating accuracy is hard. A practical approach: Maintain a fact sheet per market: leadership, headquarters, divisions, products, certifications. Flag answers that contradict it. Have a reviewer confirm and tag each issue with a GEO layer. Step 5: Dashboards Three charts are enough for leadership:
Named rate by market and engine over time
Cited rate by market and engine over time
Accuracy rate with a list of open issues by layer Resist vanity metrics. One impressive answer isn't a trend. Step 6: Close the loop Each issue should create a ticket: Layer Typical fix Owner Content Rewrite page opening with specific, self-contained facts Content / web Identity Align names, addresses and descriptions across properties Brand / ops Corroboration Correct third-party pages; earn accurate coverage PR / partnerships

Handling multiple markets and languages
Singapore-based teams often cover Malaysia, Indonesia, Thailand, Vietnam and the Philippines from one hub. A few design choices make that manageable:
One prompt file per market and language. prompts/id.yml for Bahasa Indonesia, prompts/th.yml for Thai, and so on. Don't machine-translate the English prompts; ask local colleagues how buyers actually phrase the questions.
Market-specific brand aliases. Regional entities often trade under slightly different names. Keep the alias list per market so the named check doesn't miss them.
Separate fact sheets. Leadership, addresses and product availability differ by country. Accuracy checks should run against the right market's facts.

aliases/sg.yml

market: SG
aliases:

"Example Group"
"Example Group Singapore"
"Example Holdings Pte Ltd" domains:
example.com
example.com.sg Data quality pitfalls A few things that quietly break trend lines:
1. Changing prompts mid-quarter. If you must change one, retire its ID and create a new one.
2. Mixing surfaces. API answers and consumer-app answers can differ. Log the surface in a column and don't compare across it.
3. Personalisation and location. Logged-in sessions and location can change answers. Use consistent, clean sessions for manual collection.
4. Single-run noise. AI answers vary between runs. For important prompts, collect two or three samples and record the majority result. Governance Store raw answers in a restricted dataset. They can contain inaccurate claims about your company or competitors, and you don't want them circulating out of context. Share the dashboards widely and the raw logs narrowly. Alerting Dashboards are checked monthly; problems can appear any day. Add two simple alerts: Accuracy drop: if the accuracy rate for any market falls by more than 15 points between runs, notify the brand owner for that market. New wrong fact: if a reviewer tags an answer with a fact that isn't on the fact sheet (a wrong CEO name, a closed office, a discontinued product), open a ticket automatically. Keep alerts rare and specific. A channel full of noise is a channel nobody reads. What the research supports (and doesn't) The 2024 paper that named GEO (Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi; KDD 2024) found that evidence-rich content gained visibility (up to 40%, varying by domain) and that keyword stuffing had little or negative effect. It doesn't offer a deterministic formula, and engines change frequently. Your pipeline is how you learn what works for your brand. A sensible first cycle 20 prompts × 3 engines × 3 markets = 180 answers a month Two reviewers, half a day each, for accuracy tagging One monthly report to marketing and communications leadership For teams looking at enterprise ai seo, this pipeline is the measurement backbone everything else depends on. If you'd like an outside audit before building in-house, Kiyado's team offers one free: SEO and GEO for Singapore enterprises →

AEO for SaaS Docs and Help Centres: A Step-by-Step Implementation Guide

Rida Moideen — Tue, 27 Oct 2026 16:38:07 +0000

Your docs probably answer hundreds of questions. Google probably quotes a competitor's.
This guide shows how to turn a SaaS help centre into something answer engines can lift, using the framework from Kiyado's article on GEO vs SEO vs AEO vs AIO. It's written for engineers and technical writers, with markup you can copy.
If you're comparing vendors and searching for the best agency for ai visibility services aeo geo, use this as a scope checklist. Anything an agency proposes should cover these steps.
SEO vs AEO: what actually changes
The article lists three differences:

SEO
AEO
Target
A list position
The single answer slot
Format
Thorough pages
A clean answer under a question-style heading
Measurement
Rankings and clicks
Owning the snippet or voice answer, clicks or not

And one dependency that matters for engineering: AEO depends on SEO. If a page doesn't rank, its answer can't be lifted. So we start with crawlability.
Step 1: Make the help centre crawlable
Many help centres are rendered client-side or sit on a subdomain with a restrictive robots.txt. Check both.
curl -s https://help.example.com/robots.txt
curl -s https://help.example.com/articles/reset-password | grep -i "<h1"
If the second command returns nothing, crawlers may be seeing an empty shell. Use server-side rendering or static generation for articles.
Step 2: One question per article, in the H1 or H2
Answer engines match questions to headings. Write headings the way users ask:

How do I export invoices to Xero?

Not:

Integrations > Accounting > Export

Step 3: Put the answer first
Directly under the heading, write a self-contained answer. Practitioners often suggest 40–60 words. The Kiyado article labels this a commonly cited practitioner guideline, not a published rule, so treat it as a target, not a hard limit.

How do I export invoices to Xero?

Go to Settings → Integrations → Xero and click Connect. After
you authorise access, choose which invoice statuses to sync
(for example, Paid and Sent). Invoices export automatically
every hour; use "Sync now" to export immediately. Requires the
Pro plan or above.
Then add steps, screenshots and edge cases below.
Step 4: Add structured data where it's honest
FAQ or HowTo markup can help machines understand structure. Only mark up content that's visible on the page, and keep it in sync with the text.

{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [{
"@type": "Question",
"name": "How do I export invoices to Xero?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Go to Settings → Integrations → Xero and click Connect..."
}
}]
}

Search engines change how they display rich results, so don't rely on markup alone. Clear headings and answers do most of the work.
Step 5: Name the product in the answer
Write "Example Ledger exports invoices to Xero" rather than "It exports invoices." Lifted answers lose their surrounding context, so the product name needs to be inside the answer.
Step 6: Measure AEO separately from SEO
Because AEO is measured by ownership, not clicks, set up a simple tracker:
query,ranks_page1,snippet_owner,ai_overview_cites_us,checked
"export invoices to xero",yes,competitor-a,no,2026-10-15
"reset 2fa example ledger",yes,us,yes,2026-10-15
Review monthly. Pages that rank but don't own the snippet are your best opportunities.
Step 7: Extend to AIO and GEO
Once answers are liftable, the higher layers become reachable.
AIO: check whether Google's AI Overviews cite your help articles for your priority queries.
GEO: ask ChatGPT, Gemini and Perplexity product questions ("Which invoicing tools integrate with Xero for UK agencies?") and record whether you're named.
The article explains that GEO goes further than AEO: engines combine many sources, answer "who should I use" questions and depend on off-site signals such as review platforms, directories and press. It also cites G2's 2026 research: buyers trust AI recommendations but still check review sites. Keep your G2 and Capterra listings consistent with your docs.
Step 8: Keep answers fresh with a docs review cycle
Lifted answers age badly. A snippet that still says "click the blue Export button" after a redesign actively hurts you, because the answer box keeps showing old instructions under your brand.
Add three lightweight safeguards:

Tie articles to features. Tag each help article with the feature flag or component it documents. When the component changes, the article appears in the release checklist automatically.
Add a "last verified" date to every article and review anything older than six months.
Lint the first paragraph. A simple CI check can flag articles where the first paragraph after the H1 is missing, too long or doesn't contain the product name. def check_answer(md: str, product: str) -> list[str]: body = md.split("\n", 1)[1].strip() first = body.split("\n\n", 1)[0] words = len(first.split()) issues = [] if words > 80: issues.append(f"first paragraph is {words} words") if product.lower() not in first.lower(): issues.append("product name missing from answer") return issues The 80-word threshold is deliberately loose. The goal is to catch answers that have drifted into marketing copy, not to enforce a rule the research doesn't support. Common mistakes Answers buried below marketing copy. Put the answer first. Headings that describe navigation, not questions. Outdated answers. A lifted answer with old UI steps damages trust. Separate vendors for SEO and AEO. The article warns against split retainers; both edit the same pages. Rollout plan Week 1: crawlability audit and fixes (Step 1) Week 2: rewrite the 20 most-viewed articles (Steps 2, 3 and 5) Week 3: add structured data and the tracker (Steps 4 and 6) Week 4: baseline AIO and GEO checks (Step 7) Then repeat for the next 20 articles. Within a quarter, your help centre stops being a support cost and starts being a visibility asset. A note for Manchester teams Manchester's SaaS scene competes with London and with US vendors for the same buyers. Clear, liftable answers are one of the few areas where a smaller team can out-execute a bigger competitor. It's discipline, not budget. If you'd like an outside full-stack audit, Kiyado offers one free: SEO and GEO for Manchester businesses →