The cheapest way to track AI mentions of your brand

You do not need software to find out whether AI assistants mention your product. You need software when you want to know why, and whether a change is real.

By CiteGraph · Updated 2026-08-23 · Prices checked on the vendors’ own pages

Start with the free method, because it works

The honest starting point costs nothing. Write down the ten questions a buyer would actually type — not keywords, questions: “best X for a two-person team”, “cheaper alternative to Y”, “which X handles Z”. Ask each one in ChatGPT, Claude, Gemini and Perplexity. For every answer record two things: were you named in the text, and which links did the answer cite.

That second column is the one people skip, and it is the one that matters. Being named tells you the model knows you exist. The cited links tell you which pages the answer was assembled from — and those pages are the only thing you can actually go and change.

Do this once and you will learn more about your AI visibility than any dashboard will tell you in a week. Do it three times and you will discover the harder problem: the answers move between runs, and you cannot tell whether you improved or the model simply rolled differently.

Where the free method breaks down

Answers vary run to run

Ask the same question three times and you will often get three different shortlists. A single check cannot separate a real change from variance, which means it also cannot tell you whether anything you did worked. The fix is repetition and arithmetic: sample each question several times per engine and report the rate with its margin of error, so a move only counts when it exceeds the noise.

Nobody records the citations

Copying links out of answers by hand is tedious enough that people stop after the first round. But the citation map is the actionable artefact: it tells you which third-party page decides your category, whether a competitor owns it, and whether it says anything substantive about you at all.

It does not scale past one product

An hour a week for one product is fine. For an agency with ten clients it is a day a week, every week, and the record is only as good as the person keeping it.

What the paid tools cost

Entry prices published on the vendors’ own pricing pages, checked 23 August 2026. Where a vendor does not publish a price, we say so rather than guess.

Sources: https://otterly.ai/pricing/ · https://www.tryprofound.com/pricing · https://crowdreply.io/pricing · https://www.citegraph.app/pricing. Peec AI plan names are published; prices were not rendered on its pricing page when we checked.

When it is worth paying

Pay when one of three things is true. You need to know whether a change is real, which requires repeated sampling and stated margins. You need the citation map rather than a mention count, because you intend to go and change the pages behind the answers. Or you are doing this for more than one product, where the manual method quietly becomes someone’s job.

If none of those are true yet, run the free method for a month. You will get further than most people who buy a dashboard and check it twice.

What we measured in this category

We put 10 buyer questions about AI visibility tools to 3 engines, 60 answers in total, during the week of 2026-08-17. The tools assistants named most often were Otterly (70%), Peec AI (55%) and Profound (50%). CiteGraph was named in none of them. That board is public, including the questions asked and our own absence from it: the AI search visibility leaderboard.

We publish that because a tool selling citation measurement should be willing to publish its own. If you are choosing between vendors, ask each of them what their own number is.