The 20-Point AI-Visibility Checklist
The first pass I run to see if AI engines can find, trust, and cite your site. The same one I ran across a 14-site finance portfolio.
Score each check 0, 1, or 2
Every check below has three buttons. Click one on each of the 20 checks: zero means missing, one means partial, two means solid. Your running total out of 40 and your verdict update live in the bar at the top as you go. Three questions decide everything an engine does with a page, can it find you, will it trust you, can it quote you.
Can AI find you?
Technical and crawlability. The floor. If the engine can't fetch and index a clean page, nothing else matters.
Key content is in server-rendered HTML, not built by JavaScript
Load the page with JS disabled, or use view-source. If the main content is missing, most AI crawlers see nothing. GPTBot, ClaudeBot, and PerplexityBot read raw HTML and don't run JavaScript. Google and Gemini are the exception. This one is underrated.
AI crawlers aren't blocked
Check robots.txt and firewall rules for GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, Google-Extended, and Bingbot. Many hosts block these by default and nobody notices. Robots.txt is the real control here, not llms.txt.
Indexed, with clean URLs and a current sitemap
Index coverage is the gatekeeper, because ChatGPT and Copilot lean on Bing's index and Gemini leans on Google's. No orphan pages, a real internal link path to every page, and an XML sitemap with accurate lastmod dates.
Stable for crawlers
No 4xx or 5xx errors on key pages, and a fast, reliable server response when a bot hits the site. An uncrawlable page is an uncitable page. Core Web Vitals help humans, but there's no evidence they drive AI citation, so don't spend your energy there for this.
Will AI trust you?
Entity, authority, and earned media. This is the center of gravity in 2026, and it lives mostly off your own site.
Earned media: recent third-party coverage in sources your buyers trust
The single highest-impact factor in 2026. Are you being written about, mentioned, and cited on reputable third-party sites, not just your own? If AI isn't reading you in the coverage, it isn't naming you in the answer.
Present on the platforms the engines actually lean on
Wikipedia and Wikidata for entity grounding, plus Reddit, YouTube, and the review sites in your category. Reddit is the most-cited domain across the major engines. A real, earned presence on these beats anything you can publish on your own site.
A named author who's a resolvable entity, not just a byline
Not admin, not staff. A real person with a bio, a track record, external profiles, and third-party bylines that let an engine confirm they exist.
Organization and Person schema with sameAs links
Structured data that tells engines who you are and connects your site to your other official profiles and your knowledge-graph entry. Treat this as entity clarity, not a magic citation lever. Use it to remove ambiguity, not as the strategy.
Claims corroborated across independent sources
State a fact, cite a named reputable source, and make sure the same fact shows up on other independent sites. Engines prefer claims they can safely repeat and cross-check.
Consistent entity facts and real brand-mention volume across the web
Same name, same founder, same details on Wikidata, Crunchbase, LinkedIn, and directories.
Real signals of first-hand experience
Original data, first-hand detail, and specifics a generic AI article would never contain. Adding real statistics is one of the two highest-lift on-page changes there is.
Can AI quote you?
Content structure and citability. Give the engine a clean, self-contained answer it can lift and attribute.
Answer-first structure
The direct answer sits in the first two sentences under each heading, before the throat-clearing, in about 50 to 70 words. This is the strongest on-page citation driver.
Headings phrased as the real questions people ask
The way a person actually prompts an engine. Question-format queries trigger an AI answer about 86% of the time, and your headings should match those sub-questions.
Product, Review, or entity schema with real, populated attributes
This is the one schema type with citation evidence. Skip FAQ and Article rich-result markup for this purpose, Google deprecated FAQ rich results in May 2026 and there's no citation lift from generic Article schema.
Scannable, liftable formatting
Short paragraphs, lists, and tables. Engines lift clean, self-contained passages, so give them clean units to lift.
Freshness signals that are real
A true dateModified in the schema and genuinely updated facts, not just a changed date. About 65% of AI crawler hits target content under a year old. Stale pages get skipped.
One clear topic per page
No two pages fighting over the same query. Focus aids passage extraction and entity clarity.
A topical cluster built for query fan-out
Google's AI surfaces break one query into eight to twelve sub-queries. Cover the sub-intents with a real hub-and-spoke cluster and internal links, not one keyword page.
Original assets worth citing, including video
A data study, a calculator, a comparison table, a directory, and a YouTube video. YouTube is one of the most-cited sources and often surfaces for sub-queries your text never ranks for. Give the engine something it can't get anywhere else.
You're actually measuring AI visibility
Bing Webmaster Tools for citation counts on the Copilot and Bing surface, GA4 for referral traffic from ChatGPT, Perplexity, and the rest, and a per-engine citation tracker. If you aren't tracking it, you're flying blind.
What does NOT move AI citations
Most AI SEO advice sells these three as levers. Current evidence says they aren't, at least not for getting cited. I'd rather you know.
- llms.txt. No major AI engine reads it yet. Google's John Mueller confirmed no AI system currently uses it, and 97% of llms.txt files got zero traffic. Nice idea, not a citation lever today. Ahrefs, May 2026
- Generic schema and FAQ markup. Adding JSON-LD produced no citation lift in a controlled test, and Google deprecated FAQ rich results in May 2026. Schema still helps entity clarity, but don't expect citations from it. The exception is attribute-rich Product and Review schema, check 14.
- Core Web Vitals. They matter for humans and classic SEO. There's no evidence they drive AI citation. Don't let a green score convince you that you're AI-visible.
Your verdict
Score all 20 above and your band lights up here, and at the top, live.
How AI engines actually pick who to cite in 2026
Now you've scored your site, here's the research the checklist is built on. Getting cited is a different game from ranking, and the center of gravity has moved off your own site.
- Getting cited is two gates, not one. The engine has to retrieve your page, then choose to quote it. Most retrieved pages never get cited. Being findable is necessary and nowhere near sufficient.
- Ranking on Google is no longer the same as being cited by AI. The overlap between AI-cited pages and Google's top 10 fell from 76% to 38% for AI Overviews in one half of 2025, and only about 12% of ChatGPT, Gemini, and Copilot citations rank in Google's top 10 for the prompt. Ahrefs, Dec 2025
- Earned media is the biggest lever, by a wide margin. Third-party coverage, not your own site, is where AI pulls most citations from. Earned media accounts for about 84% of all AI citations, and a University of Toronto study found a systematic bias toward earned media across ChatGPT, Perplexity, and Gemini. Muck Rack, May 2026 · Chen et al., Sept 2025
- The engines lean on a handful of platforms. Reddit appears in about 40% of AI answers and Wikipedia in about 26%. Semrush, 2026
- Freshness is a real, measured signal. About 65% of AI crawler hits target content published in the last year. Seer Interactive, Oct 2025
- On the page, concreteness wins. Adding real statistics, real quotations, and real cited sources lifts AI visibility by up to about 40%. Answer-first structure and liftable formatting are the other strong on-page levers. Aggarwal et al., KDD 2024
- Honest note on the data. Most large citation-share numbers come from vendor studies that agree with each other and with two academic papers, but aren't platform disclosures. Treat every figure as directional and dated.
Want the full picture?
This is the fast first pass. If you want the deep read, I'll run my full AI-Visibility Audit on your site, free. It scores your site across 8 dimensions and 5 AI engines, and hands you the three highest-impact fixes, whether or not we ever work together.