Free tool · AI visibility

AI visibility checklist: 21 answers, one prioritised action plan.

Answer 21 questions about how AI search and answer engines reach, read and trust your site. You get a score for each of six areas and an action plan ordered by impact and effort: why each fix matters, with its source, the exact steps, and how to check it worked. Download the plan as Markdown or CSV, print it, or copy the checklist. It runs entirely in your browser, and nothing you enter is sent anywhere.

Self-assessment, not a site scan. General guidance, with no guarantee of AI citations or rankings.

Start the checklist Read the full playbook

What the AI visibility checklist checks

AI visibility, often called generative engine optimization (GEO), is the work of making your pages easy for answer engines such as ChatGPT search, Perplexity, Claude, Copilot and Google's AI features to reach, understand and cite. Most of it is ordinary SEO done carefully. Google says there are no additional requirements to appear in AI Overviews or AI Mode beyond being indexed and eligible for Search (Google Search Central). The other engines run their own crawlers, which you have to let in.

The 21 checks fall into six areas, roughly in the order they fail:

  • Crawler access Can the AI search and answer crawlers reach your pages at all?
  • Bing and IndexNow Are you in the index that Copilot and ChatGPT search draw on, and does it hear about changes?
  • Structured data and entity Can a machine tell who you are, who wrote what, and when?
  • Content and linking Is your content easy to quote, worth quoting, current and connected?
  • Authority off your site Do people and sites that engines trust vouch for you?
  • Measurement Can you see AI referrals and citations, so you know what worked?

The tool does not fetch or scan your site. You answer from what you know, and every question says how to check if you are not sure. For the long version, read our step-by-step answer engine optimization guide and the generative engine optimization playbook, or see how our visibility optimisation service runs this work.

Answer the 21 questions

Pick the closest answer. Partly means some of it is in place. Not sure puts a check-first step into your plan. You can skip questions; the plan covers what you answer.

Used only to label your plan and, if you ask us for help, to fill in the contact form. It stays in your browser.

Crawler access

Can the AI search and answer crawlers reach your pages at all?

01Does your robots.txt let the AI search and answer crawlers in?

How to check: Open yoursite.com/robots.txt. Look for Disallow rules under OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User, Bingbot and Googlebot, and under User-agent: *.

02Have you decided, on purpose, whether AI training crawlers may use your content, and does robots.txt name them?

How to check: Look for GPTBot, ClaudeBot, Google-Extended, CCBot and Applebot-Extended in robots.txt. Where they are missing, your User-agent: * rules decide for you.

03Does your CDN or firewall let those crawlers through, without a block or a challenge page?

How to check: In Cloudflare, look at AI Crawl Control and the Bots settings. Elsewhere, check your firewall (WAF) and bot-management rules for anything that blocks or challenges AI bots.

Bing and IndexNow

Are you in the index that Copilot and ChatGPT search draw on, and does it hear about changes?

04Is your site verified in Bing Webmaster Tools, with your sitemap submitted and your key pages indexed?

How to check: Sign in at bing.com/webmasters. Look at Sitemaps, and use URL Inspection on your five most important pages.

05Do you notify search engines through IndexNow when you publish or update a page?

How to check: Look for an IndexNow key file at your site root, an IndexNow setting or plugin in your CMS, or Cloudflare Crawler Hints turned on.

Structured data and entity

Can a machine tell who you are, who wrote what, and when?

06Does your homepage carry Organization structured data, with a stable @id, your logo and sameAs links to your official profiles?

How to check: Paste your homepage URL into validator.schema.org and look for an Organization node with name, url, logo and sameAs.

07Do your articles carry Article (or BlogPosting) markup with a named Person as author, and dates that match the page?

How to check: Run one post through validator.schema.org. Look for Article or BlogPosting, an author that is a Person with a url, and datePublished and dateModified.

08Do pages that answer buyer questions have a visible FAQ, with FAQPage markup that matches it word for word?

How to check: Open a service page or a popular post. Is there a visible FAQ? Compare its questions and answers with the FAQPage JSON-LD in the page source.

09If you publish an llms.txt file, is it current, and does it list your key pages?

How to check: Open yoursite.com/llms.txt and compare it with your current key pages.

10Do your name, legal entity, address, founders and service descriptions match across your site, LinkedIn, directories and other profiles?

How to check: Put your About page next to your LinkedIn company page, Crunchbase, Google Business Profile (if you have one) and your main directory listings. Compare names, spelling, addresses, founders and services.

Content and linking

Is your content easy to quote, worth quoting, current and connected?

11Do your key pages open each section with a direct one- or two-sentence answer, before the detail?

How to check: Read the first two sentences under each heading on your top five pages. Do they answer the heading on their own?

12Do you publish original data (a survey, benchmark, index or dataset) with its method, sample size and date?

How to check: List any page on your site that contains numbers nobody else has. Does it state how, when and from how many the data was collected?

13Do your statistics and claims link to the named primary source: the study, the regulator, the vendor's documentation?

How to check: Pick a key page and list every number on it. Does each one link to the original source that states it?

14Do important pages show a visible last-updated date that changes only with real edits, and are they reviewed when facts change?

How to check: Open your top guides and comparison pages. Is there a visible date, does it match dateModified in the markup, and when did the content last really change?

15Does every important page get contextual links from related pages on your own site?

How to check: Run a site crawler or your SEO tool's site audit and look for important pages with zero or one internal link pointing to them.

Authority off your site

Do people and sites that engines trust vouch for you?

16Does your content name a real author, with an author page, relevant credentials and a LinkedIn link?

How to check: Open three posts. Is there a named author, does the name link to an author page, and does that page show experience and a LinkedIn profile?

17Do you have genuine profiles and reviews on the directories and review sites your buyers use?

How to check: Search for your company name plus reviews, and look at the review and directory sites that appear for your category (for software, sites such as G2).

18Are your company or your experts mentioned and quoted on third-party sites buyers trust: LinkedIn, trade press, podcasts, community answers?

How to check: Search your company and founder names in news and on LinkedIn for the last six months. Count mentions you did not publish on your own site.

Measurement

Can you see AI referrals and citations, so you know what worked?

19Does your analytics separate visits from AI assistants (ChatGPT, Perplexity, Claude, Gemini, Copilot), and do you review them monthly?

How to check: In GA4, open Reports, Acquisition, Traffic acquisition and look at session source for chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com.

20Do you check Bing Webmaster Tools' AI Performance report and Google Search Console every month?

How to check: When did someone last open AI Performance in Bing Webmaster Tools and the Performance report in Search Console, and write down what they saw?

21Do you run a fixed set of buyer questions through ChatGPT, Perplexity, Gemini, Claude and Copilot at least quarterly, and check they describe you accurately?

How to check: Is there a saved list of buyer questions, with dated results per engine, from the last 90 days?

Scoring needs JavaScript. Without it, every action is in the full playbook below, with why it matters, how to do it and how to check it.

Your scores and action plan

Your scores and action plan appear here as you answer. Scoring runs in your browser with JavaScript; without it, use the full playbook below, in default priority order.

Have Leapbuzz do this for you

Send your plan to a senior practitioner. We fill in our contact form with your top gaps and your website (with JavaScript off, the link opens the empty form); you review the message, add your details and send it. Nothing leaves this page until you submit that form.

Have Leapbuzz do this for you

Prefer email? Write to [email protected]. We reply within one business day.

The full playbook: all 21 actions

Every action the plan can include, grouped by area, in default priority order within each area. Evidence labels: CONFIRMED means a primary source (the vendor, Google, a regulator or a peer-reviewed paper) states it; REPORTED means a credible secondary source such as an industry study or trade press, so treat it as directional. Sources were reviewed on 2 October 2026.

Crawler access

Allow the AI search and answer crawlers in robots.txt

Do first if it is a gapImpact HEffort L

These crawlers decide whether a page can appear in ChatGPT search, Claude search and Perplexity answers. OpenAI recommends allowing OAI-SearchBot for search inclusion, Anthropic says blocking Claude-SearchBot prevents indexing for search, and Perplexity says PerplexityBot is the crawler that surfaces and links sites. While they are blocked, nothing else in this plan can work.

How to do it

  1. Open robots.txt and list every User-agent group with its Disallow lines.
  2. Remove Disallow rules that cover your key pages for OAI-SearchBot, Claude-SearchBot, PerplexityBot, Bingbot and Googlebot, and for the user-triggered fetchers ChatGPT-User and Claude-User. If a User-agent: * group blocks everything, add an explicit group for each of these agents with Allow: /.
  3. Perplexity-User and ChatGPT-User fetch pages when a person asks, and their vendors say robots.txt may not apply to them; for those two, your CDN setting (the next action) is what decides access.
  4. Keep private areas (admin, cart, staging, internal search) disallowed for everyone.
  5. Add a Sitemap: line with the full sitemap URL on your canonical host, for example https://yoursite.com/sitemap.xml.
  6. Publish the file, then allow about a day: OpenAI says robots.txt changes take about 24 hours to reach its search.

How to check it worked

  • Open the live robots.txt in a private window: each agent above has either no group of its own (so User-agent: * applies) or a group with no Disallow covering your key pages.
  • Google Search Console's robots.txt report shows the version Google fetched, with no errors.

Answers question 01: Does your robots.txt let the AI search and answer crawlers in?

Stop your CDN or firewall from blocking AI search crawlers

Do first if it is a gapImpact HEffort L

An Allow in robots.txt means nothing if the CDN answers the crawler with a block or a challenge page. Cloudflare began blocking AI crawlers by default for new zones on 1 July 2025, so many sites block them without knowing it.

How to do it

  1. In Cloudflare, open AI Crawl Control and the Bots settings, including any block-AI-bots option and Bot Fight Mode. Allow the search and answer crawlers (OAI-SearchBot, ChatGPT-User, Claude-SearchBot, Claude-User, PerplexityBot, Perplexity-User).
  2. Check firewall custom rules and rate limits for user-agent, country or network rules that catch these crawlers.
  3. On other CDNs and hosts, find the equivalent AI-bot or bot-management setting; managed hosting plans often have one.
  4. Set the training crawlers to match your robots.txt decision.

How to check it worked

  • In your CDN's bot or crawler analytics, requests from OAI-SearchBot, Claude-SearchBot and PerplexityBot get 200 responses, not 403s or challenges.
  • A curl request with a bot user agent is only a rough test: some CDNs verify bots by IP address, so the analytics view is the real check.

Answers question 03: Does your CDN or firewall let those crawlers through, without a block or a challenge page?

Decide your AI training policy and name the training agents

Quick win if it is a gapImpact MEffort L

Training and search are separate switches. GPTBot is OpenAI's training crawler and is independent of OAI-SearchBot; ClaudeBot is Anthropic's. Google-Extended is a robots.txt control token rather than a separate crawler: it does not affect Google Search, but it decides whether Gemini Apps and Vertex AI can ground answers on your pages, so blocking it to protect content also takes you out of Gemini grounding.

How to do it

  1. Decide with whoever owns content and legal risk: allow training (more presence in model knowledge) or block it (more control over reuse).
  2. Write one explicit group per training agent (GPTBot, ClaudeBot, CCBot, Applebot-Extended, and the Google-Extended token) with Allow: / or Disallow: /.
  3. Name the current agents from each vendor's crawler page. Do not rely on old or unofficial names such as anthropic-ai, Claude-Web or User-agent: Gemini to carry your decision; an old line can stay, but it does not replace the current names.
  4. Add a comment in robots.txt with the decision and its date, so nobody reverses it by accident.

How to check it worked

  • Each training crawler has exactly one group in the live robots.txt, and it matches the decision.
  • The search and answer crawlers from the previous action are still allowed.

Answers question 02: Have you decided, on purpose, whether AI training crawlers may use your content, and does robots.txt name them?

Bing and IndexNow

Get indexed in Bing and use Bing Webmaster Tools

Quick win if it is a gapImpact HEffort L

Bing's index feeds Microsoft Copilot, and Bing is a named search provider behind ChatGPT search. OpenAI also runs its own index, so Bing is necessary, not sufficient. Bing Webmaster Tools also has an AI Performance report with Copilot citations, cited pages and grounding queries.

How to do it

  1. Add and verify your site at bing.com/webmasters. You can import a verified Google Search Console property.
  2. Submit your XML sitemap on your canonical host.
  3. Use URL Inspection on your most important pages and request indexing for any that are missing.
  4. Fix the crawl problems it reports: pages blocked by robots.txt, redirect chains, 4xx and 5xx errors.

How to check it worked

  • The Sitemaps report shows your sitemap processed, with URLs discovered.
  • URL Inspection shows your key pages as indexed, and a site:yoursite.com search on bing.com lists them.

Answers question 04: Is your site verified in Bing Webmaster Tools, with your sitemap submitted and your key pages indexed?

Send IndexNow pings for new and changed pages

Quick win if it is a gapImpact MEffort L

IndexNow tells participating engines, including Bing, that a URL has changed, so updates reach Bing's index, and the AI answers built on it, sooner. Google does not use IndexNow, so keep your sitemap as well. Cloudflare Crawler Hints can send the pings for you.

How to do it

  1. Turn on Cloudflare Crawler Hints, or your CMS's IndexNow setting or plugin, or add a deploy step that posts changed URLs to the IndexNow endpoint with your key.
  2. Host the key file at your site root as the IndexNow documentation describes.
  3. Ping only the URLs that really changed. Do not resubmit the whole sitemap on every deploy.

How to check it worked

  • Your key file loads at https://yoursite.com/your-key.txt.
  • Bing Webmaster Tools lists IndexNow submissions for your recent changes.

Answers question 05: Do you notify search engines through IndexNow when you publish or update a page?

Structured data and entity

Publish complete Organization structured data

Quick win if it is a gapImpact MEffort L

Structured data that matches your visible content helps machines resolve who you are. Google says its AI features need no special markup, so treat this as entity clarity, not a ranking trick. It is cheap, and sameAs ties your site to the profiles engines cite most, such as LinkedIn.

How to do it

  1. Add one JSON-LD Organization node to the homepage with @id (for example https://yoursite.com/#organization), name, url, logo, and sameAs listing your LinkedIn page and other official profiles.
  2. Add legal name, address and contact details only where they match what the site shows.
  3. Reference the same @id everywhere the organisation appears, for example as the publisher of articles.
  4. Put the markup in the page HTML your server sends.

How to check it worked

  • validator.schema.org shows one Organization node and no errors.
  • Name, logo and links in the markup match the visible site and your LinkedIn page.

Answers question 06: Does your homepage carry Organization structured data, with a stable @id, your logo and sameAs links to your official profiles?

Make your entity facts identical everywhere

Quick win if it is a gapImpact MEffort L

Engines build an answer about a company from many sources, and most citations go to third-party sites such as LinkedIn rather than to brand sites. Conflicting names, addresses or descriptions make you harder to resolve. The effect on citations is plausible but weakly measured, and the fix is cheap.

How to do it

  1. Write a one-page fact sheet: official name and spelling, legal entity, registration number, address, founders, a one-sentence description, your services and your official URLs.
  2. Update the About page, footer, Organization markup and every profile (LinkedIn, Crunchbase, directories, Google Business Profile if you qualify) to match it.
  3. Choose one canonical host, with or without www, and redirect the other to it.
  4. Do not create Wikipedia or Wikidata entries unless you meet their notability rules.

How to check it worked

  • Comparing the fact sheet with each profile finds no differences.
  • Asking two AI assistants what your company does and where it is based returns the fact-sheet facts.

Answers question 10: Do your name, legal entity, address, founders and service descriptions match across your site, LinkedIn, directories and other profiles?

Mark up articles with Article and Person

When you can if it is a gapImpact LEffort L

Google's guidance is that structured data must match the visible content, and that its AI features need no special markup. Consistent author and date markup still makes authorship and freshness machine-readable at almost no cost.

How to do it

  1. Add Article or BlogPosting JSON-LD to each post: headline, author (a Person with name, a url to an author page, and sameAs to LinkedIn), publisher (your Organization @id), datePublished, dateModified and image.
  2. Make the visible byline and visible dates say exactly what the markup says.
  3. Change dateModified only when the content really changes.

How to check it worked

  • validator.schema.org on three posts: no errors, and the author is a Person.
  • On each post, the visible byline and dates equal the markup.

Answers question 07: Do your articles carry Article (or BlogPosting) markup with a named Person as author, and dates that match the page?

Keep real FAQs on the page, with markup that matches

When you can if it is a gapImpact LEffort L

Google stopped showing FAQ rich results on 7 May 2026, so FAQ markup no longer earns a search feature. A visible question-and-answer block is still easy for readers and answer engines to quote, and any markup must match the visible text.

How to do it

  1. Collect the questions buyers really ask, from sales calls, support tickets and Search Console queries.
  2. Answer each in 40 to 80 words, direct answer first.
  3. If you keep FAQPage JSON-LD, generate it from the same text so the two can never drift.
  4. Do not spend time adding FAQ markup to win a rich result; it will not show.

How to check it worked

  • Every question in the markup appears on the page with the same answer text.

Answers question 08: Do pages that answer buyer questions have a visible FAQ, with FAQPage markup that matches it word for word?

Treat llms.txt as optional: keep it accurate or leave it

When you can if it is a gapImpact LEffort L

Google says its AI features need no AI text files, special markup or machine-readable files beyond normal SEO, so llms.txt is not a lever for Google. If you already publish one, keep it accurate: a stale file is worse than none.

How to do it

  1. If you have one: generate it from your sitemap or CMS at build time, with absolute URLs and a one-line description per key page.
  2. If you do not: leave it until the higher-priority actions in this plan are done.

How to check it worked

  • Every URL in llms.txt returns 200 and matches a live page.

Answers question 09: If you publish an llms.txt file, is it current, and does it list your key pages?

Content and linking

Link related pages to each other

Quick win if it is a gapImpact MEffort L

Google's AI features guidance points site owners to the usual SEO best practices, including internal linking. Pages that nothing links to are found and crawled less, and are harder for any engine to place in context.

How to do it

  1. Group pages into topics: one main page per service or theme, with the posts that support it.
  2. Link each supporting post to its main page and to two or three related posts, with descriptive link text.
  3. When you publish a post, link to it from at least two existing pages.
  4. Fix or redirect broken internal links.

How to check it worked

  • A crawl shows no important page with fewer than two internal links pointing to it.
  • No internal link returns a 404.

Answers question 15: Does every important page get contextual links from related pages on your own site?

Rewrite your key pages answer-first

Big bet if it is a gapImpact HEffort M

Answer engines lift self-contained passages. In the GEO study (KDD 2024), adding citations, quotations and statistics raised visibility in generative-engine answers by up to 40%, with effects varying by domain. Answer-first, fact-dense sections are the format those methods produce. It was a lab benchmark, not live ChatGPT or Google, so treat the number as directional.

How to do it

  1. Pick the 5 to 10 pages that matter most to buyers: service pages, top posts, comparison pages.
  2. Under each heading, write a one- or two-sentence answer that stands alone, then the detail.
  3. Phrase headings as the questions buyers ask.
  4. Use short lists and tables for steps, comparisons and specifications, so each fact sits in a passage that can be lifted on its own.
  5. Put specific numbers, names and dates into the answer sentence where you have them.

How to check it worked

  • Read each section's first two sentences on their own: they answer the heading.
  • Note which of your test prompts cite the page before and after (see the prompt-panel action).

Answers question 11: Do your key pages open each section with a direct one- or two-sentence answer, before the detail?

Cite named primary sources next to your claims

Big bet if it is a gapImpact HEffort M

In the GEO study, citing sources and adding quotations were among the methods that raised visibility in generative answers. Named, checkable sources also protect you: Google says AI-assisted content should be fact-checked by people.

How to do it

  1. List every statistic and factual claim on your top pages.
  2. Link each one to the primary source (the study, the regulator, the vendor's own documentation), not to a blog that quotes it.
  3. Name the source and its date in the sentence itself.
  4. Remove claims you cannot source.

How to check it worked

  • Every number on the page links to a primary source that states that number.

Answers question 13: Do your statistics and claims link to the named primary source: the study, the regulator, the vendor's documentation?

Publish one piece of original data, then repeat it

Big bet if it is a gapImpact HEffort H

Unique facts are the main reason an engine cites a small site over a large publisher, and brand-owned sites are a minority of AI citations. Adding statistics was one of the methods that raised visibility in the GEO study.

How to do it

  1. Choose a question in your niche that buyers ask and nobody answers with numbers.
  2. Collect data you can defend: a customer survey, anonymised totals from your own work, or a structured review of public sources.
  3. Publish the findings on one page with method, sample size, dates and limitations, plus a chart and a plain table.
  4. Repeat on a fixed schedule, quarterly or yearly, so the dataset stays current.
  5. Never present synthetic or estimated numbers as survey results.

How to check it worked

  • The page states the method, the sample size and the collection dates.
  • Your test prompts on the topic start citing the page (check monthly).

Answers question 12: Do you publish original data (a survey, benchmark, index or dataset) with its method, sample size and date?

Keep important pages fresh, with honest dates

When you can if it is a gapImpact MEffort M

Industry studies report a freshness bias in AI Overviews and Perplexity citations, and Google's rater guidelines treat changing a date without changing the content as low quality. Real updates help; cosmetic date bumps can hurt.

How to do it

  1. Show a visible last-updated date on guides, comparisons, and pages about rules or prices.
  2. Set a review cycle (for example quarterly) and a list of triggers, such as a law, price or product change.
  3. When you update, change the content, the visible date and dateModified together, and say what changed.
  4. Never change a date without a real change.

How to check it worked

  • On each important page, the visible date equals dateModified in the markup.
  • The lastmod values in your sitemap differ between pages, rather than one build date for every URL.

Answers question 14: Do important pages show a visible last-updated date that changes only with real edits, and are they reviewed when facts change?

Authority off your site

Show who wrote each page and why they know

Quick win if it is a gapImpact MEffort L

Google's helpful-content guidance asks whether it is clear who created the content and what expertise they bring. Real author pages also give engines a person to connect with your organisation.

How to do it

  1. Create an author page per writer: name, role, relevant experience, credentials you can prove, and a link to LinkedIn.
  2. Link the byline on every post to that page.
  3. Add the author to the Article markup as a Person with the same url and sameAs.
  4. For regulated topics, add a reviewer line naming a qualified reviewer.
  5. List only true credentials.

How to check it worked

  • Every post has a linked byline, and each author page loads and links to LinkedIn.

Answers question 16: Does your content name a real author, with an author page, relevant credentials and a LinkedIn link?

Build genuine directory profiles and reviews

Quick win if it is a gapImpact MEffort L

Citation studies report that answer engines cite review and directory sites, and that Perplexity cites G2 often. Fake, AI-written or undisclosed insider reviews are banned in the United States under the FTC's 2024 rule.

How to do it

  1. List the three to five directories and review sites that appear when you search your category and your brand.
  2. Claim or create profiles with exactly the facts on your fact sheet.
  3. Ask real, recent clients for reviews, with no incentive tied to a positive rating.
  4. Reply to reviews, good and bad.

How to check it worked

  • Each profile matches your fact sheet.
  • Every review comes from a real client and is dated.

Answers question 17: Do you have genuine profiles and reviews on the directories and review sites your buyers use?

Earn mentions where answer engines look

Big bet if it is a gapImpact HEffort M

Studies of AI citations report that Reddit, Wikipedia, YouTube and LinkedIn dominate, and that for professional queries LinkedIn is the most cited domain across six AI platforms. Brand-owned sites are a minority of citations, so presence on third-party sites matters.

How to do it

  1. Publish useful, specific posts from your founder's or experts' own LinkedIn profiles each week, written by them.
  2. Answer journalist requests on expert-quote platforms with a specific insight that can be attributed to you.
  3. Pitch a few relevant podcasts, then turn each episode into a post and short clips.
  4. Answer genuine questions in communities under a disclosed identity, and link only when it is the best answer.
  5. Never use fake accounts, bought engagement, or automation tools that break platform rules; LinkedIn's User Agreement bans automated posting, commenting and messaging outside its approved APIs.

How to check it worked

  • Each month, count new third-party mentions and links (search alerts or a mentions tool).
  • Your test prompts start to name you, or to cite pages that mention you.

Answers question 18: Are your company or your experts mentioned and quoted on third-party sites buyers trust: LinkedIn, trade press, podcasts, community answers?

Measurement

Track AI assistant referrals as their own channel

Quick win if it is a gapImpact MEffort L

AI answers send few clicks, so the visits you do get are easy to miss. Pew Research Center found that people clicked a link inside a Google AI summary on about 1% of visits, and clicked any result less often when a summary appeared (8% against 15%). Give AI assistants their own channel, watch direct visits to deep pages (much AI traffic arrives with no referrer), and ask leads how they found you.

How to do it

  1. In GA4, create a custom channel group with a channel called AI assistants, placed above Referral, matching session source by this regex: chatgpt\.com|openai\.com|perplexity\.ai|claude\.ai|gemini\.google\.com|copilot\.microsoft\.com|bing\.com/chat|you\.com|meta\.ai|deepseek
  2. In other analytics tools (Cloudflare Web Analytics, Plausible, Matomo), save the same referrer filter.
  3. Add a How did you hear about us? question to your lead forms, with an AI assistant option.
  4. Review monthly. Numbers are small and swing, so compare quarters, not weeks.

How to check it worked

  • The AI assistants channel shows sessions in a 90-day report.
  • New leads answer the How did you hear about us? question.

Answers question 19: Does your analytics separate visits from AI assistants (ChatGPT, Perplexity, Claude, Gemini, Copilot), and do you review them monthly?

Review Bing AI Performance and Search Console monthly

Quick win if it is a gapImpact MEffort L

Bing Webmaster Tools' AI Performance report (public preview since February 2026) shows Copilot citations, cited pages and the grounding queries that found you. In Google Search Console, AI Overviews and AI Mode traffic is counted inside the Web search type, so watch your key pages there.

How to do it

  1. In Bing Webmaster Tools, open AI Performance and export cited pages and grounding queries every month.
  2. In Search Console, track impressions and clicks for your key pages in the Web search type, and use any generative AI reporting your property shows.
  3. Turn the grounding queries into content briefs: the questions that found you, and the near misses.
  4. Keep a simple monthly log: date, citations, top cited pages, notes.

How to check it worked

  • A dated monthly log exists, with the last three months filled in.

Answers question 20: Do you check Bing Webmaster Tools' AI Performance report and Google Search Console every month?

Run a fixed prompt panel every quarter

When you can if it is a gapImpact MEffort M

No single report shows AI answers across engines: Bing's AI Performance report covers Copilot only. A fixed panel of buyer questions, run on a schedule, shows whether engines name you, whether they describe you correctly, and which third-party sources they cite instead. It is a measurement method, not a ranking lever, and answers vary from run to run.

How to do it

  1. Write 20 to 50 questions real buyers ask: about your brand, your category, and comparisons.
  2. Run each one in ChatGPT, Perplexity, Gemini, Claude and Copilot with web search on, three times each, from a clean or logged-out profile where possible.
  3. Log whether you are named, whether the description is accurate, and which domains are cited.
  4. Fix inaccuracies at the source (your site, LinkedIn, directories), then run the panel again next quarter.

How to check it worked

  • A dated spreadsheet holds every prompt, engine and run, with the cited domains.

Answers question 21: Do you run a fixed set of buyer questions through ChatGPT, Perplexity, Gemini, Claude and Copilot at least quarterly, and check they describe you accurately?

How the plan is ordered

Each action has an impact rating (high, medium or low) and an effort rating, taken from our October 2026 review of vendor documentation and industry studies. Your answer sets the size of the gap: No counts in full, Not sure counts three quarters and adds a check-first step, Partly counts half, and Yes moves the action to your keep-it-that-way list.

Priority score = 10 × impact (high 3, medium 2, low 1) × gap ÷ effort (low 1, medium 2, high 3), rounded. Crawler access comes first whenever it is not fully open: those two actions get 100 extra points, because no other fix matters while the crawlers are blocked. Ties keep the order of the questions. Each action in your plan and in the downloads shows its score and how it was worked out.

Area and overall scores weight each answer by its action's impact: Yes scores full marks, Partly half, No and Not sure zero. One exception: having no llms.txt file scores three quarters, because the evidence says it matters little.

Frequently asked questions

What is an AI visibility checklist?

A list of checks on whether answer engines such as ChatGPT search, Perplexity, Claude, Copilot and Google's AI features can reach your pages, understand who you are, and find something worth citing. This one turns your answers into a scored action plan, ordered by impact and effort, with a source, steps and a way to verify each fix.

Does the tool scan or fetch my website?

No. It is a self-assessment. Your answers and the website name you type stay in your browser, and this page sends nothing anywhere. Each question includes a short how-to-check note, so you can answer from evidence rather than memory.

Which AI crawlers should robots.txt allow?

For search and answers: OAI-SearchBot and ChatGPT-User (OpenAI), Claude-SearchBot and Claude-User (Anthropic), PerplexityBot and Perplexity-User (Perplexity), plus Bingbot and Googlebot. The user-triggered fetchers may not follow robots.txt, so check your CDN settings too. Training is a separate decision: GPTBot, ClaudeBot and the Google-Extended token. Blocking Google-Extended does not affect Google Search, but it stops Gemini Apps and Vertex AI from grounding answers on your pages.

Does llms.txt help my site get cited?

Not as far as public evidence shows. Google says its AI features need no AI text files or special markup. If you already publish llms.txt, keep it accurate; if you do not, the other actions in your plan matter more.

Is FAQ schema still worth adding?

Not for a search feature: Google stopped showing FAQ rich results on 7 May 2026. Visible questions with direct answers are still useful to readers and easy for answer engines to quote. If you keep FAQPage markup, make it match the visible text exactly.

How do I see traffic from ChatGPT and Perplexity?

Create an AI assistants channel in your analytics that matches referrers such as chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com. Review Bing Webmaster Tools' AI Performance report for Copilot citations, and ask leads how they found you, because much AI traffic arrives with no referrer.

Will following the plan get my brand cited by AI assistants?

There is no guarantee. Engines do not fully publish how they choose sources, and they change it without notice. The plan removes known blockers and follows what primary documentation and industry studies support, which improves your odds, but nobody can promise citations or rankings.

General guidance, no guarantee

This checklist is general guidance based on public documentation and industry research reviewed on 2 October 2026. It is not a site audit, and it does not guarantee citations in ChatGPT, Perplexity, Claude, Gemini, Copilot or Google's AI features, or any search ranking. Engines change how they select sources without notice, so check the linked sources before acting on a specific claim.