Free AI Visibility Score: What the Number Actually Means

A free score compresses many separate answers into one number. Three things get averaged, and each one hides detail.

• Engines. ChatGPT, Perplexity, Gemini, Claude and Google's AI features are different products. Google's own documentation says AI Overviews and AI Mode "may use different models and techniques" and can return different links.
• Prompts. The tool picks a few questions. The engine often doesn't search for those exact words. OpenAI says ChatGPT search "typically rewrites your query into one or more targeted queries." So any prompt set is a proxy for queries nobody outside the engine can see.
• Search-on and search-off answers. As of October 2026, ChatGPT decides per query whether to search, and Claude.ai search is a user toggle. A search-off answer comes from training data frozen at a cutoff. OpenAI's model docs list a February 16, 2026 cutoff for the GPT-5.6 family and August 31, 2025 for the Instant path. Nothing you publish this quarter reaches a search-off answer.

So a score of 40 might mean one engine names you every time and three never do. For the mechanics behind the number, see how an AI visibility score is calculated.

Two tools can score the same brand 70 and 25 on the same day, and both can be honest. They differ on which engines they ask, which prompts they run, how they weight each engine, and what counts as a hit.

Even the unit differs by engine. Anthropic's web search tool cites short text spans, while other engines cite whole sources. There is no shared denominator to average across.

Formula versions break comparisons too. Say a tool subtracts a fixed penalty for every wrong fact with no denominator. A bigger prompt set then finds more errors in absolute terms, and the score falls even when the error rate stays flat. When a vendor fixes a formula like that, old and new scores stop being comparable.

No outside ground truth exists to settle the disagreement. Google's generative AI report in Search Console shows impressions, with no clicks and no position. Any "AI rank" a tool reports has nothing to be checked against.

The next move: compare a tool only with itself, on the same prompts and the same formula version. If a tool doesn't say which version produced a score, ask.

You can run a rough version of any free check in about an hour.

1. Write three buyer questions. Use what a customer would ask before buying, not your brand name. Say you run a dental clinic in Austin: "emergency dentist in Austin open on Sunday."
2. Use a clean session. Log out, or turn memory off. OpenAI says ChatGPT may use saved memories when rewriting a search query, which makes results personal and hard to repeat.
3. Ask each question on each engine. Run them in ChatGPT, Perplexity, Gemini and Claude, then in Google Search to see whether an AI Overview appears.
4. Record four things per answer. Were you named? Were you linked or cited? Were the facts right (price, location, product names)? Did the answer show sources, meaning it searched?
5. Check that search crawlers can reach you. Open yourdomain.com/robots.txt. Rules are per bot. OpenAI's crawler page says GPTBot governs training and OAI-SearchBot governs ChatGPT search, and each setting is independent. A file can allow search while blocking training:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

6. Test the firewall. Request a page with the crawler's published user-agent string, for example curl -I -A "<OAI-SearchBot string from OpenAI's bots page>" https://yourdomain.com/. A 403 or a challenge page means your CDN blocks a bot that robots.txt allows.
7. Save it with a date. Repeat monthly with the same questions.

The do-it-yourself AI visibility audit expands each step.

Start with the answers, not the number.

Look engine by engine. In OtterlyAI's 2,000 AI blogs experiment, 2,718 of the 3,934 tracked citations for one test site (69%) came from Microsoft Copilot, while ChatGPT, Gemini, AI Overviews and AI Mode almost never cited it. A blended score would have hidden that split. The trade-offs are laid out in one score vs per-platform scores.

Treat a zero carefully. A zero on three prompts means AI didn't name you for those questions, in that session. It doesn't prove AI has never heard of you. What a zero score means covers the usual causes.

Weigh wrong facts above missing mentions. An answer that names you with an old price can do more harm than one that skips you. A free AI hallucination check focuses on that.

Watch the trend. A move from 20 to 35 on the same prompts and formula tells you something. A single 35 against a competitor's 60 from another tool tells you nothing.

A free score is a snapshot, and a small one.

• Sample size. Three prompts can't represent every way buyers ask. In Otterly's test, Microsoft's own AI Performance report counted 42,817 citations for the same site where the prompt tracker logged 3,934. The time windows may differ, but any prompt sample is a sample. See how many prompts you need.
• Hidden queries. You see your prompt, not the queries the engine actually ran.
• Cited is not the same as used. OpenAI says the sources a model consulted often outnumber the ones it cites.
• No promises. No one controls what a model says. You do control whether crawlers can reach your pages, whether key facts sit in the HTML before JavaScript runs, and whether your site states those facts consistently.

If you report the score upward, the playbook on explaining score limits to stakeholders helps set expectations.

The free checker on this site runs the first half of the manual method for you. It asks up to three buyer questions, either yours or ones written from your homepage, and scores the answers with the 99Visibility score. Treat it like the hand check: open each engine's answer, note who was named and which facts were wrong, and record the date.

It is a snapshot, not tracking. If you want the same prompts rerun over time on the same formula, you can start on the 99Visibility Free plan with no card.

Is a free AI visibility score accurate? It accurately reports what happened for the prompts it ran, on the engines it asked, at that moment. It isn't a complete picture of your AI visibility. The prompt set is small, engines rewrite queries, and some answers come from training data rather than live search. Read it as a snapshot and check the answers behind it.

Why did my score change when nothing on my site changed? Engines change models and defaults, and they decide per query whether to search. An answer that searched last month may come from training data this month, or the reverse. Your CDN can also change what crawlers may reach without your edits. Before drawing conclusions, confirm the tool used the same prompts and the same formula version both times.

Can adding schema or an llms.txt file raise my score? The evidence doesn't support it. Ahrefs' matched-control study of 1,885 pages found that adding schema produced no major citation uplift, and AI Overview citations fell 4.6%. Google says Google Search ignores llms.txt, and no other engine documents reading it. Existing markup does no harm, but it isn't a lever for AI answers.

Does a higher AI visibility score mean more traffic? Not necessarily. A score counts mentions or citations in answers, not visits. Google's generative AI report in Search Console shows impressions only, with no clicks, so even Google's own data can't connect AI visibility to traffic. Check your analytics for AI referrals separately, knowing that some AI surfaces send no trackable referral at all.

How often should I check my free AI visibility score? Monthly is enough for most brands. Use the same three buyer questions each time, a logged-out or memory-off session, and the same tool. Check again soon after any site change that affects access, such as a new CDN, firewall rules, robots.txt edits or a move to client-side rendering.