GEO tools: what they do and how to choose

What generative engine optimization tools do, the four types worth knowing, how to evaluate paid ones and which free checks to run first.

The short answer

A GEO tool measures or checks something. None of them can make an AI engine cite you. The useful ones tell you whether engines can reach your pages, whether your markup is valid and whether your brand shows up in answers. The rest is writing pages worth citing.

Be careful with tools that promise rankings or citations. Google’s guide to generative AI features in Search says that no third-party tool has access to its internal ranking or AI systems, and it advises using such tools only if they help your workflow, while judging their advice against Google’s official guidance. That is a sound rule for any engine.

Four types of GEO tool

1. Technical checkers. They answer one question: can engines fetch and understand this page? They look at robots.txt, rendering, canonical URLs, noindex tags, hreflang and structured data.

  • Free options: the URL Inspection tool in Google Search Console, the Rich Results Test, the Schema Markup Validator at validator.schema.org and Lighthouse.
  • Our free AI visibility audit checks title, description, canonical, hreflang, JSON-LD, meta robots and AI crawler access from pasted HTML. It runs in your browser and sends nothing to a server.

2. Official performance reports. For Google, the generative AI performance report in Search Console shows how people find your content through Google’s generative AI features in Search and Discover. It is first-party data, which no third-party estimate can match for Google.

3. Prompt and visibility trackers. These paid products run a set of questions across AI engines on a schedule and log whether your brand is mentioned and your URLs are cited. They are the only way to see non-Google engines at scale, but they come with limits:

  • Answers vary between runs, so one result means little. Trends across many prompts matter.
  • Results change with location, language and the model version in use.
  • Citation detection is imperfect. A good tool shows you the raw answer so you can check it.

4. Analytics and server logs. Segment referral traffic from AI products in your analytics, and look in server logs for AI crawler user agents such as GPTBot, ClaudeBot and PerplexityBot. This tells you what actually reached your site.

How to evaluate a paid tool

Ask these questions before paying:

  1. Which engines and which versions does it query, and from where? Coverage that does not match your audience’s languages and countries is of little use.
  2. How many times does it run each prompt? One run per prompt is noise.
  3. Can you bring your own prompts? Generic prompt lists rarely match your customers’ questions.
  4. Can you see and export the raw answers? If you cannot verify a citation, you cannot trust the score.
  5. What exactly does the score measure? Be wary of single proprietary numbers with no definition.
  6. Does it claim access to hidden ranking data? Treat that as a red flag.
  7. What does it cost over a year? Many tools price by prompt volume, so costs grow with your prompt set.

I do not rank vendors here. Feature sets and prices change quickly, and a list would be out of date soon. The questions above stay useful.

What you can do with a spreadsheet

You can start without paying anything:

  1. Write 20 to 50 questions your customers actually ask, in each language you serve.
  2. Run them monthly in ChatGPT, Perplexity, Google AI Overviews and Claude.
  3. Record whether your brand is named, whether your URL is cited and which competitors appear.
  4. Repeat each prompt a few times and note how much the answers differ.

This is slow, but it shows you what a paid tracker will automate, and it gives you a baseline to compare against.

What not to buy

  • Guaranteed citations or rankings. Nobody can guarantee them.
  • “Secret” Google metrics. See the point above about third-party tools.
  • Bulk content for every query variant. Google’s guide says that producing separate pages for each variation of how people might search, mainly to manipulate rankings or generative AI answers, violates its spam policy on scaled content abuse, and that more pages do not make a site better.
  • Special files presented as required. The same guide says Google Search does not use llms.txt files or other special machine-readable AI files. Other systems may read them, so they are harmless, but they are not a ranking lever for Google.

Source: Google Search Central, guide to optimizing for generative AI features in Google Search.

Where to start

  1. Run the free audit on your key pages and fix what it flags.
  2. Check Search Console, including the generative AI report, if you target Google.
  3. Build your prompt set and record a baseline.
  4. Only then decide whether a paid tracker would save you time.

If you would rather have an expert review all of this, the GEO audit covers it and ends with a prioritized report. For the concepts behind each check, read the GEO guide.