Does a zero AI visibility score mean AI doesn't know my brand?

A visibility score counts how often your brand shows up in answers to a fixed list of prompts. A zero means none of those answers named you. That is a narrow fact about one sample.

Three things shape that sample:

• The prompts. A tool asks the questions it chose. Your buyers may ask different ones.
• The retrieval state. Engines decide per query whether to search the web. ChatGPT search is model-discretionary, so some answers come from live results and some from the model's frozen training.
• The date. Models update and indexes change. A result from one day can differ the next.

Any single score, including ours, averages answers given with and without live retrieval. That is why AI visibility scores average more than they appear to. A zero tells you these prompts did not surface you. It does not tell you the model has never heard of you.

The model's knowledge cutoff

When an engine answers without searching, it relies on what it learned before its cutoff. OpenAI's model docs (verified August 2026) list a February 16, 2026 cutoff for the GPT-5.6 family and August 31, 2025 for the Instant path, chat-latest. Anything you launched after those dates cannot appear in an answer given with search off. If you're a young company, read whether ChatGPT can know a new company.

Blocked access

With search on, the engine can only cite pages its crawler could fetch. OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. GPTBot is a separate setting that governs training only. A firewall or CDN can also return a 403 to a crawler that robots.txt allows. And Vercel found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so content that only appears after scripts run can be invisible to them.

Prompt choice

OpenAI says ChatGPT search rewrites your prompt into one or more targeted queries. Google may fan out into related searches. Nobody outside the engine sees those real queries, so every tracked prompt is a proxy. Broad category prompts also tend to favor well-known brands, since the evidence shows a strong association between brand prominence and citation.

Run this test in the engine that scored you zero. Use a logged-out session or turn memory off, because OpenAI notes saved memories can change how ChatGPT rewrites a query.

1. Ask a direct branded question with search off. For example, "What does [your brand] do and what does it cost?" Use a setting or API call with web search disabled. This shows what the model knows from training alone.
2. Ask the same question with search on. Note whether it finds your site and cites it.
3. Ask three or four questions your buyers really ask, in their words, with search on.

If the model says it has nothing on you, see why ChatGPT says it has no information about your company.

• One platform at zero, others not. Engines use different crawlers and indexes. A blended score can hide a single blocked bot, so read each platform on its own.
• Google results. Google states a page must be indexed and eligible to show with a snippet to appear in AI Overviews or AI Mode, and that AI Overviews often do not trigger at all. Some tools label a Gemini API call as "Google AI", which is a different product. See why a tool's Google AI result may not match the AI Overview you see.
• Small samples. Three prompts is a snapshot. One extra prompt can move the result from zero to something, or back.

Start with what you control. Confirm robots.txt lets OAI-SearchBot, PerplexityBot and Claude-SearchBot in. Then confirm your firewall doesn't turn them away: the AI Crawler Checker requests a page as each AI crawler and shows which ones get blocked. Check that your pricing and core claims are in the HTML before JavaScript runs.

Next, state specific facts (a price, a number, a one-line definition) on pages crawlers can fetch. That gives an engine a sentence it can quote. Where engines cite other sites in your category, being named on those sites is associated with appearing in answers, with no guaranteed effect. For early-stage brands, see how a startup with no press gets mentioned.

What you can't control: the model's training data, its cutoff, and which retrieved pages it chooses to cite. Re-run the free checker on this site after each fix, on more than one date. If you report to a board, explain the score's limits to stakeholders before the number gets treated as a ranking.

Does blocking GPTBot cause a zero in ChatGPT? Not for ChatGPT search. OpenAI states its crawler settings are independent of each other. GPTBot controls whether content may be used for training. OAI-SearchBot controls whether your site can appear in ChatGPT search answers. A site can block GPTBot and still be eligible for search, as long as OAI-SearchBot is allowed and your firewall lets it through.

Will adding schema markup move my score off zero? The evidence doesn't support that. Google says structured data isn't required for its generative AI features. Ahrefs tracked 1,885 pages that added schema against matched controls and found AI Overview citations fell 4.6%, with no major lift elsewhere. Existing markup does no harm, but it isn't a lever for AI answers. Access and clearly stated facts are where the evidence points.

Can a zero score still hide wrong information about my brand? Yes. A score that counts mentions says nothing about accuracy. When you ask a direct branded question, the model may describe you with an old price, a defunct product or a competitor's details. That is a separate problem from visibility. Note each wrong fact, then make sure the correct version is stated plainly on a page crawlers can fetch.

Is the free checker on this site enough to judge my visibility? It's a useful first look, not a verdict. The free checker on this site runs up to three buyer questions and scores them once. It is a snapshot, not continuous tracking. Treat a zero as a reason to run the search-off and search-on test above, then check crawler access, rather than as proof that AI doesn't know you.