Back to Blog

There is a lot of writing about AI search tools. AEO platforms, GEO trackers, LLM visibility monitors, ChatGPT rank checkers. What there is almost none of is writing about the category itself: who publishes it, what shape the results take, and whether any of it can actually earn a click.

So we measured it. On August 1, 2026 we queried Google in the United States for 20 terms covering the category, captured the top 20 organic positions for each, and classified every result. That is 330 organic results. The full dataset, including our domain classification, is published below so anyone can re-cut it.

Every query in the category returns an AI Overview

This is the finding that reframes everything else. Not most queries. Not the informational ones. All twenty. There was no term in this category where Google returned a plain list of blue links.

The rest of the SERP furniture is heavy too. Related searches appeared on 20 of 20 queries, People Also Ask on 9, video carousels on 8, and Perspectives on 6. On a typical result in this category, the first organic listing sits well below an AI-generated answer, a question accordion, and often a video row.

The practical consequence: ranking in this category does not buy clicks. It buys impressions. Any agency or vendor measuring their AI-search content by click-through is measuring against a ceiling the SERP layout has already set. We say this as a publisher in this space, not as an observer of one.

Half the category is published by the vendors being ranked

Just over half of everything ranking in this category is published by a company selling a product in it. Nearly a third of all results are a vendor's own ranked list, which in practice means a list that includes the vendor, usually at position one or two of the list.

This is not an accusation of bad faith. Category-defining content is a legitimate and effective play, and vendors have the product knowledge to write it. But it does mean a buyer researching AI visibility tools is reading, more often than not, a ranking authored by one of the ranked parties. That is worth knowing before you trust the ordering.

One caveat we want to be explicit about. Deciding whether a domain is a vendor in this category is a judgment call, unlike the AI Overview count which is purely mechanical. Our full classification list ships with the dataset. If you disagree with a call, re-cut it and the numbers will move.

Nobody owns this category

For a category this commercially valuable, the concentration is remarkably low. The ten most frequent domains hold barely a quarter of the results. Two thirds of the domains that rank appear on exactly one query and never again.

The reflex when you see Semrush, Ahrefs and a wall of funded SaaS names on page one is to conclude the category is closed. The distribution says otherwise. This is a category with a very long tail, which means entry is possible. It is the click economics, not the competition, that make it hard.

The format everyone assumes wins does not

Conventional advice says original testing beats listicles. In this dataset it does not. The tested format averages a marginally better rank and a marginally worse top-ten share. The difference is inside the noise.

What is true is that tested pieces are scarce. Six percent of the category versus fifty-two percent listicles. If you go looking for a format advantage you will find anecdotes: a handful of "I Tested 40 Tools" pieces sitting at positions two and three. We nearly published that conclusion. Then we counted the ones that did not make it, and the advantage disappeared.

Format is not the lever here. Over half the category is reviewing tools in one of two shapes, and neither shape is winning on its own merits.

What this means if you are publishing here

Three things follow from the data.

First, set the right success metric. With an AI Overview on 100% of queries, clicks are not the outcome this category produces. Impressions, brand presence inside AI answers, and inbound interest from other publishers are. If your reporting judges this content on click-through you will kill work that is doing its job.

Second, do not compete on format. Listicle versus tested is a coin flip. What is genuinely scarce is subject: 58% of the category is reviewing tools and almost nobody is analyzing the category itself. That gap is wide open, and it is the one gap a vendor cannot fill without indicting their own listicle.

Third, expect a long tail rather than a landgrab. Two thirds of ranking domains appear once. Getting in is realistic. Owning the category is not, at least not through content alone.

Method

Twenty queries covering the AI search tools category, run against Google in the United States on August 1, 2026, top 20 organic positions captured per query, 330 organic results total after excluding SERP features. AI Overview presence was recorded mechanically from the SERP response. Format classification uses a published regular expression against result titles. Vendor classification is a manual judgment call and the full domain list is included in the dataset.

Limitations worth stating. One market and one day, so this is a snapshot rather than a trend. Twenty queries is enough to be confident about a 100% and 51.5% finding, and not enough to split those numbers finely by sub-topic. Google personalizes and localizes results, so your own SERPs will differ at the margins. The dataset is published precisely so you can check ours.

Download the full 330-row dataset (CSV) or read the method record (JSON).

Questions

Do AI search tool queries always show an AI Overview?

In our sample, yes. All 20 queries we tested across the AI search tools category returned an AI Overview on Google in the United States on August 1, 2026. There were no exceptions in the dataset.

How much of the AI search tools category is written by the vendors themselves?

We classified 51.5% of the 330 organic results as published by a company selling a tool in that same category, and vendors held 53.2% of top ten positions. Roughly 31.2% of all results were a vendor-authored "best of" list.

Is the AI search tools category too competitive to enter?

The concentration says no. We found 172 unique domains across 330 results, and 66.3% of those domains appeared exactly once. The ten most frequent domains held only 27.6% of results. Entry is realistic. The difficulty is that AI Overviews suppress clicks, not that competitors block ranking.

Do original "we tested" articles outrank listicles?

Not in this dataset. Listicles averaged rank 11.3 with 47.7% reaching the top ten. First-person tested pieces averaged rank 10.0 with 45.0% reaching the top ten. The difference is within noise, so format is not a reliable lever here.

What should you measure if you publish content in this category?

Impressions, presence inside AI answers, and inbound interest from other publishers. With an AI Overview on every query in our sample, click-through is capped by SERP layout, so judging this content on clicks will cause you to cut work that is performing.

Can I see the underlying data?

Yes. The full 330-row dataset, the list of queries, and our vendor domain classification are published as CSV and JSON so the analysis can be independently re-cut.