GEO Checklist: 40 Checks for AI Search Visibility
A GEO checklist is a structured self-audit of whether AI engines can find, understand, trust, and cite your brand. Below are 40 checks across five pillars. Tick what is already true, count your score, then fix the gaps in the order the pillars appear: each one depends on the one before it.
If any term below is unfamiliar, the AI search glossary defines it, and the complete guide to generative engine optimization explains the framework behind the checks.
1. Access and rendering
Nothing else on this list matters if an engine cannot fetch and read your pages. These eight checks confirm the door is open.
- Retrieval crawlers are allowed. robots.txt permits OAI-SearchBot, ChatGPT-User, PerplexityBot, Bingbot, and Googlebot, unless you have a deliberate reason to block one. Syntax is in robots.txt for AI crawlers.
- Your CDN or firewall is not silently blocking AI bots. Several CDNs now offer AI bot blocking, and Cloudflare began blocking known AI crawlers by default for new domains in July 2025. Check the dashboard setting, not just robots.txt.
- Training crawlers are a deliberate decision. GPTBot, Google-Extended, and CCBot are allowed or blocked on purpose, with the reason written down. The tradeoff is in should you block AI crawlers.
- Key content is in the initial HTML. View source, not the browser inspector, on your top pages. The main text should be there without JavaScript running.
- Your important pages are indexed in Google. Search Console’s page indexing report shows them as indexed.
- Your site is indexed in Bing. Bing Webmaster Tools shows indexed URLs. ChatGPT search leans heavily on Bing’s index, so a site missing from Bing is structurally hard for ChatGPT to cite.
- New URLs are pushed, not just waited for. Sitemaps are submitted in both consoles and IndexNow notifies Bing and other participating engines when pages change.
- Core Web Vitals pass on mobile. LCP within 2.5 seconds, INP within 200 milliseconds, CLS within 0.1. Details in Core Web Vitals explained.
2. Entity clarity
An engine that is unsure what you are will hedge, or describe a competitor instead. These checks make your brand an unambiguous entity.
- One page says exactly what you are. Your homepage or About page states your category, services, location, founding year, and founders in plain declarative sentences.
- Organization schema with a stable
@id. Defined once, referenced from every other node, never redefined. See Organization schema for AI search. - A short, genuine
sameAslist. Only official profiles. What belongs there is covered in the sameAs guide. - Person entities for founders and authors. Real bios and credentials, linked to the organization. See Person schema and author entities.
- Your facts are identical everywhere. Name, address, phone, category, and description match across your site, profiles, and directories.
- No unresolved name collision. If another company, person, or place shares your name, you have a plan from entity disambiguation.
- Service pages carry Service schema. Each service is its own node whose provider is your organization. Template in our Service schema guide.
- Your Knowledge Panel, if one exists, is claimed and accurate. See how to claim and verify it.
3. Quotable content
AI engines quote passages, not pages. Only about 15% of pages ChatGPT retrieves end up cited, according to an AirOps analysis reported by Search Engine Land in March 2026, and structure is a large part of what separates the two groups.
- Every key page leads with the answer. The first sentence under each heading states the claim.
- Sections are self-contained. Each passage makes sense lifted out on its own, with the subject named rather than referred to as “it”. See content chunking for AI.
- Headings are the questions buyers ask, not internal labels like “Overview” or “Pricing”.
- Claims carry their evidence in the same passage. Figure, source, and date together.
- Commercial questions are answered on the page. Pricing model, timelines, who it suits, and who it does not.
- Comparison content exists. Honest X vs Y and alternatives pages for your category, since validation queries are the hardest to be cited for.
- Topic clusters cover the question space. One pillar per core topic, with supporting pages for the sub-questions. See topic clusters that AI engines cite.
- Important pages were reviewed in the last six months and show an honest updated date.
4. Off-site corroboration
Engines trust what independent sources say about you more than what you say about yourself. In 2026 correlation analyses, branded mentions tracked AI Overviews visibility roughly three times more closely than backlinks did.
- Review profiles are complete and active on the platforms your buyers use: Google, G2, Clutch, Trustpilot, or your vertical’s equivalent.
- You appear in category roundups that rank in both Google and Bing.
- Directory listings are complete and consistent with your canonical facts.
- A Wikidata item exists, if you meet notability, with every statement sourced. See Wikidata for brands.
- Independent publications describe you accurately. Recent press, podcasts, or guest pieces that name you and your category.
- Founders have a visible, consistent presence under their own names.
- You have original data or research someone could cite.
- Community presence is genuine. Real participation by people who know the field, never manufactured posts. The playbook is in digital PR for AI search.
5. Measurement
You cannot manage AI visibility you do not measure, and traffic alone undercounts it because most AI answers are zero-click.
- A fixed prompt set of 30 to 50 real buyer questions.
- Prompts run across ChatGPT, Perplexity, Gemini, and Google AI Mode.
- Each prompt is run several times per cycle, because answers vary run to run.
- Presence, citations, accuracy, and competitor mentions are logged every cycle. Method in measuring brand visibility in ChatGPT.
- AI referral traffic is segmented in GA4. Setup in tracking AI traffic in GA4 and Search Console.
- Search Console and Bing Webmaster Tools AI reports are checked where your account has access.
- Share of model is tracked against named competitors. Defined in share of model.
- The cadence is monthly and has a named owner.
Score yourself
Count your ticks. This is a rough guide rather than a benchmark, but the bands map to fairly predictable situations.
| Score | What it usually means | Focus next |
|---|---|---|
| 32 to 40 | You are ahead of most of your category | Keep the cadence; expand clusters |
| 20 to 31 | Foundations are in place | The gaps are usually off-site and measurement |
| Under 20 | Engines may struggle to find or identify you | Pillars 1 and 2 before anything else |
What to fix first
| Order | Pillar | Why it goes here | Typical effort |
|---|---|---|---|
| 1 | Access and rendering | Gates everything else | Low to medium |
| 2 | Entity clarity | Decides whether you are recognized | Medium |
| 3 | Quotable content | Decides whether retrieved pages get cited | Medium to high |
| 4 | Measurement | Start early so you have a baseline | Low |
| 5 | Off-site corroboration | Slowest to build, so begin in parallel | High |
Measurement jumps ahead of off-site work for one reason: a baseline taken after you have changed things cannot tell you what worked. Capture it before your first fix ships.
Popular GEO advice we left off this checklist
The GEO checklists ranking alongside this one share a lot of good advice, and several items above appear in all of them. A few popular recommendations are missing here on purpose. Each one is either unproven or quietly counterproductive.
| Common advice | Why it is not on this list | What to do instead |
|---|---|---|
| Put a 40 to 60 word answer under every heading | The length is a rule of thumb, not a measured threshold, and forcing it makes pages read like a template | Answer in the first sentence under each heading, at whatever length the answer needs |
| Update timestamps every 30 to 90 days | Changing the date without changing the content is a false freshness signal, and engines can compare versions | Refresh on a schedule, and move dateModified only when the facts or guidance actually change |
| Publish an llms.txt file | No major AI search engine has confirmed that it reads one when choosing sources | Spend the time on server rendering and Bing indexation; see what is llms.txt |
| Turn every heading into a question | Question headings help when they match a real question and hurt when they are forced | Use a question where a buyer would ask one, and a plain claim elsewhere |
| Add FAQ schema to every page | FAQ markup only helps when the questions and answers are visible and genuinely asked | Mark up FAQs that exist on the page for readers; see FAQ schema for AI answers |
The test we apply to every item is simple: would it still be worth doing if AI search did not exist? Everything on the 40-point list passes it. The items in this table mostly do not.
Where the GEO checklist fits in your AI visibility work
This checklist tells you where you stand; it does not tell you what AI engines currently say about you or which competitors they name instead. For that, run your prompt set, or request our free AI visibility audit: we run the prompts across the major engines, score these five pillars, and hand back a prioritized fix list. For the deeper technical layer, continue with the technical SEO checklist for AI search.
Frequently asked questions
What is a GEO checklist?
A GEO checklist is a structured self-audit of the things that decide whether AI engines like ChatGPT, Perplexity, Gemini, and Google AI Mode can find, understand, trust, and cite your brand. It covers technical access, entity clarity, content structure, off-site corroboration, and measurement.
How often should I run a GEO checklist?
Run the full checklist quarterly and after any major site change such as a redesign or migration. Run the measurement section monthly, because AI answers shift faster than your site does.
Which GEO checks matter most?
Access and indexation come first, because nothing downstream works if an engine cannot fetch or find your pages. Entity clarity comes second, because an engine that is unsure what you are will describe a competitor instead.
Can I complete a GEO checklist without an agency?
Yes. Most checks are things a marketer or developer can verify in an afternoon. The harder parts are sustained off-site work and keeping a measurement cadence going month after month.
Should I update my page dates regularly for GEO?
Only when the content genuinely changes. Freshness matters to AI engines, but a new date on unchanged content is a false signal and can be detected by comparing versions. Review your cited pages on a schedule, change what is out of date, and let dateModified move with those real changes.
How long until GEO changes show up in AI answers?
Retrieval-side fixes such as access, indexation, and page structure can show up within days to weeks. Entity and off-site work compounds over months, because it depends on independent sources and, eventually, on new model versions.