Free AI Visibility Audit: What It Should Check and What the Free Plan Covers

A good audit answers three questions, in this order of control.

• Can AI crawlers reach and read your pages? This is fully in your control, and the evidence behind it is the strongest in the field.
• What do AI engines say about you? Wrong prices, old product names and confused competitors are specific, fixable problems.
• Do engines name you for the questions buyers ask? This is the outcome everyone wants, and the one you control least.

Most free tools only cover the third question. That leaves you with a number and no fix. If you want the longer do-it-yourself version, our step-by-step audit method goes deeper on each part, and the AI brand audit guide covers how engines describe a company.

You need a spreadsheet, a terminal and about two hours.

1. Write five buyer questions. Use phrasing a customer would type, such as "best payroll software for a 20-person agency". Avoid your brand name in at least three of them.
2. Ask each engine, logged out. Run every question in ChatGPT, Perplexity, Claude and Gemini with memory off. OpenAI says ChatGPT may use saved memories when it rewrites a search query, so logged-in results won't reproduce. Record whether you were named, who was named instead, and which URLs were cited.
3. Note the retrieval state. If an answer shows no citations, assume it may come from the model's training data, not a live search. Retrieval is discretionary on every platform as of October 2026.
4. List your true facts and compare. Write down your pricing, founding year, headquarters, product names and integrations. Mark every answer that states one of them wrong. Then check your own site: the same fact stated two ways on two pages is a common cause.
5. Read your robots.txt per bot. Open https://yourdomain.com/robots.txt. OpenAI documents that each of its bots is controlled independently, so this is valid:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

GPTBot governs training only. OAI-SearchBot controls ChatGPT search. Also check PerplexityBot, Claude-SearchBot, Googlebot, bingbot and Meta-WebIndexer. 6. Test what the server returns. Run curl -sI -A "OAI-SearchBot" https://yourdomain.com/pricing and look at the status line. A 403 or a challenge page means something is blocking that user agent. Your laptop isn't on OpenAI's IP ranges, so your CDN or server logs are the real check. 7. Check your facts are in the raw HTML. Run curl -s https://yourdomain.com/pricing | grep -i "per month". Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them. If your price only appears after JavaScript runs, those crawlers can't read it. 8. Look for suppression tags. Search the page source for noindex, nosnippet, max-snippet:0 and data-nosnippet. Google and Bing document these as controls that keep content out of AI answers. Templates often apply them by accident.

Sort every finding into one of three buckets.

• Broken access. A blocked retrieval bot, a 403 from the firewall, a key fact missing from the raw HTML, or an accidental nosnippet. Fix these first. They are binary, and the fix is yours to make.
• Wrong facts. Note the engine, the question, the wrong claim and any cited source. If the source is your own outdated page, update it. If it is a third-party page, that page is the thing to correct.
• Missing mentions. Look at which sites the engines cited instead of you. Being named on those sites is associated with appearing in the answer, though no study promises an effect.

Keep results per engine. One engine can name you often while another never does, and an average hides both facts. Our page on what a free visibility score means explains why a single number needs that breakdown.

Be plain with yourself about the limits.

• Your prompts are a proxy. OpenAI says ChatGPT search rewrites a question into one or more targeted queries. You never see those, so your five questions approximate real demand.
• It is one snapshot. Answers change with model updates and retrieval decisions. One run cannot show a trend. See what a snapshot misses.
• Citation is not influence. OpenAI's API returns more retrieved sources than cited ones, so a missing citation doesn't prove your page went unused.
• There is no AI rank. Google's own reporting for AI features shows impressions only, with no position data.
• Nothing guarantees inclusion. Google states a page must be indexed and snippet-eligible to appear in AI Overviews, and that serving still isn't guaranteed.

The critical survey of 45 GEO studies (arXiv, July 2026) found no technique with a stable, cross-platform causal effect on discoverability. An audit finds blocks and errors. It does not produce outcomes.

The 99Visibility Free plan costs $0 and runs 1 audit a month on 5 keywords for 1 brand, across 2 AI engines. Paid plans audit 5 engines. Each audit reports how those engines describe the brand, flags wrong facts, checks whether AI crawlers can reach the site, and lists recommended fixes. It covers steps 2, 4 and part of 5 to 7 above without the spreadsheet.

The free checker on this site is a lighter version: up to three buyer questions, scored once, with no account. If you want the monthly audit, you can start on the Free plan with no card. Either way, treat the output as a list of things to fix, not a forecast.

Work through findings in order of control.

1. Fix access: allow the retrieval bots you want, confirm your CDN returns 200 to them, and remove accidental suppression tags.
2. Put load-bearing facts (price, plans, a one-line definition of what you do) in the server-rendered HTML. A specific, fetchable sentence gives an engine something it can quote.
3. Make each fact identical across every page that states it.
4. Contact third-party sites that state wrong facts about you.
5. Re-run the same questions a month later and compare per engine.

For platform-specific checks, the ChatGPT visibility checklist walks through access to accuracy for one engine. If you are weighing this against your SEO audit, how the two differ covers the overlap.