AEO tools: the five categories, and what each one actually measures

There is no single answer engine optimisation tool. What is sold as AEO or GEO tooling falls into five categories that measure different things: answer visibility trackers that run prompt sets across assistants, AI modules added to established SEO platforms, crawler and log checkers that tell you whether AI fetchers can reach your site, page and markup tools, and plain referral analytics. Most teams need one tracker plus two free things they already have. This page says what each category measures so you buy the one that answers your question.

The five categories

Category	What it measures	Free option that covers part of it
Answer visibility trackers	How often you are named across a prompt set, in which engines, with which sources cited	A spreadsheet of thirty prompts, run by hand
AI modules in SEO platforms	Brand mentions in AI answers, next to your existing rank and keyword data	Search Console for Google surfaces
Crawler and log checkers	Whether named AI agents reached your pages, and what they received	Your own server or CDN access log
Page and markup tools	Structure, schema validity, answer placement, llms.txt presence	View source, the Rich Results Test, Lighthouse
Referral analytics	Visits arriving from assistant domains	Your existing analytics, with a referrer filter

Answer visibility trackers

These are the products built specifically for this job. They hold a prompt list, run it across several assistants on a schedule, and report how often you and your competitors are named, plus the domains cited in each answer. Products in this category include Profound, Peec AI, Otterly.AI and Scrunch AI, and the category is growing quickly enough that any list is a snapshot.

The five categories above tell you what each tool watches. One measurement sits outside all of them: whether a buyer asking a plain question gets your name back. AI Knows Us takes that reading, without putting the brand in the question, and then does the page and listing work behind whatever it finds. It is worth a look before you compare paid options, because it answers the question the category labels do not.

What to check in any of them: how many engines, how many runs per prompt, whether you can see and edit the prompt list, whether the raw answer text is stored, and whether you can set the country and language. We benchmarked six of them side by side in AI visibility tools compared, and tested the agency reporting versions in white label AI visibility reports.

AI modules inside SEO platforms

The established platforms have added answer visibility features to what they already sell. Ahrefs has Brand Radar, Semrush has its AI Toolkit, and BrightEdge reports on AI answer presence alongside its existing data. Check your current subscription before you buy anything, because the feature may already be included.

The advantage is real: one seat, one set of keyword and rank data, and your AI numbers in the same report as your organic numbers. The thing to confirm before you rely on it is the measurement design behind the number, which is the same question you would ask of a dedicated tracker. How many engines, how many runs, and can you see the prompts.

Crawler and log checkers

This category is unglamorous and it is where most problems are actually found. If OAI-SearchBot has never fetched your pricing page, no amount of writing fixes your ChatGPT visibility.

• Your access log is the primary source. Filter for GPTBot, OAI-SearchBot, ChatGPT-User, PerplexityBot, ClaudeBot and Google-Extended and look at both the hit counts and the status codes.
• Your CDN dashboard. Cloudflare, Fastly and similar products have bot management rules that can refuse unfamiliar agents before your server ever sees them.
• A crawler such as Screaming Frog or Sitebulb, set to a named user agent, shows you what each agent receives, including pages that render only after JavaScript.

The decisions to make here are in which AI crawlers should you allow.

Page and markup tools

These check the page rather than the answer: whether the direct answer is near the top, whether headings are the questions, whether your structured data is valid and matches the visible text, and whether you have an llms.txt file. Google's Rich Results Test and Lighthouse cover most of it at no cost, and Lighthouse now audits for an llms.txt file as part of its agentic browsing checks.

Worth keeping in proportion: Google states in its AI features documentation that no new machine readable file or special markup is required to appear in AI Overviews or AI Mode. Use these tools to keep the page clean, and put the real effort into what the page says.

Referral analytics

Add a filter to your analytics for traffic from assistant domains, including chatgpt.com, perplexity.ai, claude.ai and gemini.google.com. The volumes are usually small. The sessions tend to be further along than organic ones, because the person arrived after reading a recommendation. This costs nothing and almost nobody does it.

How to choose

• Start from your question. "Are we named?" needs a tracker. "Why are we not named?" needs the log checker and a page review first.
• Count your prompts. Under about twenty prompts in two engines, a spreadsheet and an hour a month is a real measurement. Above that, the arithmetic buys the tool.
• Insist on a visible prompt list. A score you cannot trace to a question is a number you cannot act on.
• Ask for runs per prompt. Several runs per prompt is what turns a single reading into a measurement you can compare month to month.
• Check the engines against your buyers. If your market lives in Perplexity and the tool covers ChatGPT only, the coverage gap outweighs every feature.
• Check locale support if you sell in more than one country, because location changes which companies appear.
• Check the exports. You want the raw answers out of the tool, so your figures survive a change of supplier.

Vendor pricing in this category is published on each product's own site and changes often, so check it there rather than relying on any list.

This page is about tool categories. For the measurement design that any tool should implement, see how to report AI search visibility. For traditional audit tooling, see SEO audit tools, what each kind actually checks.

Questions buyers ask about AEO tools

Do I need an AEO tool at all?

Not on day one. Thirty prompts, three engines, three runs each and a spreadsheet gives you a defensible baseline. Buy a tool when the monthly collection stops fitting in the time you have.

Can my existing SEO platform do this?

Several now report AI answer presence, so check your current subscription before buying anything new. Confirm the engines covered and the runs per prompt, then compare that with a dedicated tracker on the same prompt list.

Why do two tools report different numbers for me?

Because they are running different prompts, different numbers of runs, different engines and different locales. Line up those four and the gap usually explains itself. This is why the prompt list matters more than the headline score.

Is there a free AEO tool?

The free stack is Search Console for Google surfaces, your access log for crawler reach, your analytics for assistant referrals, and a spreadsheet for the prompt runs. That combination answers most of the questions a small business has.

What about llms.txt generators?

They save you twenty minutes of typing. The file is short enough to write by hand, and the writing is the part that matters, because the summary line and the per link notes are what make it useful.

What to do first

• Write your prompt list before you look at any product. The list is the asset; the tool is the way you run it.
• Check your access log for the named AI agents. Fix access before buying measurement.
• Add an assistant referral filter to your analytics today, since it is free and takes minutes.
• Trial two trackers on the same prompt list, in the same week, and compare the raw answers rather than the scores.
• Confirm you can export the raw answers before you commit to anything.