Which GEO tool tells you how to fix your AI visibility, not just track it?

A tracking tool tells you whether ChatGPT, Perplexity or Gemini named you. A fixing tool tells you why it didn't and what to change. The difference shows up in a single recommendation. Read one and ask whether you could verify it yourself in five minutes.

The evidence supports two kinds of fix.

• Access fixes. These remove blocks: server-side rendering of key content, per-bot robots.txt permission, firewall and CDN reachability, index eligibility, and snippet controls such as noindex and data-nosnippet. They are binary and mechanically checkable.
• Accuracy fixes. AI platforms get specific facts wrong about specific brands. A tool that shows which fact is wrong, on which platform, and which page the answer drew on gives you something to correct.

Everything else on the market is weaker. Most published GEO techniques were measured on pages already inside the model's context, not on whether a page gets found in the first place. When you compare platforms, use the criteria in what to look for in a brand accuracy monitoring platform and hold every recommendation to the same test: finding, page, change.

Here are three findings a fixing tool should be able to surface, with the fix each one implies.

A 403 for OAI-SearchBot

The finding: your robots.txt allows OAI-SearchBot, but your CDN returns 403 when that crawler requests your pricing page. OpenAI's crawler documentation says sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. OpenAI's help centre adds that eligibility needs both things: robots.txt permission and a host or CDN that allows traffic from OpenAI's published searchbot IP addresses.

The fix: allow the verified crawler at the firewall, then retest. The robots.txt side looks like this:

User-agent: OAI-SearchBot
Allow: /

Don't confuse it with GPTBot. GPTBot governs training only, and OpenAI states each setting is independent of the others. A good tool writes this up so engineering can act on it, as in developer tickets for AI access fixes.

A price that only appears after JavaScript runs

The finding: the raw HTML of your pricing page has no prices, and the rendered page does. Vercel found that ChatGPT and Claude crawlers fetch JavaScript files (11.50% and 23.84% of requests) but don't execute them, so they can't read client-side rendered content. Googlebot does render, so this matters most for non-Google engines.

The fix: server-render the price table so it exists before JavaScript runs. The full workflow is in fixing AI answers that get your pricing wrong.

A wrong fact on a third-party page

Say you run a software company, and Perplexity's answer gives the wrong founding year while citing a directory listing. The finding is the wrong fact, the engine, and the cited page.

The fix: correct the listing at its source, and state the fact once, consistently, on a fetchable page of your own site. When engines cite other sites instead of yours, being named correctly on those sites is associated with appearing in the answer, though no effect is guaranteed. Track each error and its correction in an AI error log so you can see whether answers change.

Third-party pages can also carry injected instructions aimed at AI systems. That is a reputational risk to watch for, never a tactic to copy.

These are the most common recommendations in GEO tools, and the primary sources contradict them.

• Schema. Ahrefs tracked 1,885 pages that added schema against matched controls and found no major uplift on any platform, with AI Overview citations falling 4.6%. Google says structured data isn't required for generative AI search, and FAQ rich results were deprecated in May 2026.
• Rewriting for AI. Google's AI optimization guide says you don't need to write in a specific way for generative AI search, and it rejects chunking as a requirement. C-SEO Bench found most conversational SEO methods were largely ineffective and frequently hurt document ranking.
• llms.txt. Google says Search ignores it, and no other engine documents reading it.
• E-E-A-T scores and domain authority. Google says E-E-A-T isn't a specific ranking factor. Domain Authority is a Moz metric no engine computes.

Don't overcorrect. Google says existing markup does no harm, so ripping out schema isn't a fix either. The honest line is that it isn't a lever for AI answers. If a tool's top recommendation is on this list, ask what evidence sits behind it.

Be clear with yourself about the parts nobody controls.

• Parametric knowledge. When an engine answers without searching, it relies on what the model learned before its training cutoff. Nothing you publish this quarter reaches that answer.
• Retrieval decisions. Each engine decides per query whether to search, and rewrites your prompt into its own queries. Your fix makes a page eligible. It does not make it selected.
• Position in the answer. C-SEO Bench found a source's position in the model's context matters more than content edits, and no publisher can set that position.

So any tool promising citations, rankings or a target score is overselling. A blended visibility score, including any vendor's, averages answers given with and without live search. What you control is access, rendering and the accuracy of facts on pages engines can fetch. A good tool keeps its recommendations inside that boundary.

Run the tool on your own site and grade what comes back.

1. Start free. Several tools offer a first audit with no sales call; see which tools let you run a free audit without a demo. The free checker on this site shows whether AI engines name you for up to three buyer questions, as a snapshot.
2. Verify one finding by hand. If the tool says a crawler is blocked, request the page as that crawler. 99Visibility's free AI Crawler Checker requests a page as each AI crawler and shows which ones the firewall turns away.
3. Check it scores each engine separately. One engine can cite you heavily while others barely do, and a single number hides that.
4. Count the generic advice. If schema, llms.txt or "AI-friendly writing" lead the list, the tool is selling tactics, not fixes.
5. Retest after you ship. Access fixes are binary, so you can confirm them. Answer changes take longer and aren't guaranteed.

For a fuller feature checklist, see AI SEO audit tools: features that matter.