ResearchJuly 23, 20266 min read

By Loren Castillo, Founder, Outpace Media

We asked six AI engines to recommend a tool in our own category. Here is what they read first.

48 answers across ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude and Grok. The engines barely read company websites. They read Reddit, YouTube and a small set of list posts, and we finished last in our own study.

We build a tool that tells businesses whether AI assistants recommend them. So we pointed it at ourselves, in our own category, and let it run.

The result was not flattering, and that is the useful part.

What we ran

We gave SearchVis eight buying questions, the kind a marketer actually types when they are shopping for this sort of software:

• What is the best tool to track if ChatGPT mentions my brand?
• How do I know if Perplexity recommends my company over competitors?
• What software tracks whether Google's AI Overviews cite my website?
• Is there a tool that explains why ChatGPT cites my competitors instead of me?

And four more like them.

Each question went to six engines: ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude and Grok. That is 48 complete answers, collected in July 2026, in US English. For every answer we stored the full text, every source the engine cited, and every brand it named.

It is a small sample and a single category. Treat the exact percentages as directional and the pattern as the finding.

Finding 1: the engines are not reading company websites

Here are the sources cited most often, counted as the number of the 48 answers that cited each domain at least once.

Reddit is the single most cited domain in the category, ahead of every vendor, every review site and every SEO publication. YouTube is second.

The long tail matters more than it looks. Roughly forty different blogs each showed up in two to five answers, and almost all of them were the same shape of page: a numbered list post titled some variation of "the best AEO tools" or "how to track brand mentions in ChatGPT". No single one dominates. Collectively they are where the answer comes from.

One technical note for anyone running a similar study. A large share of Gemini's citations point at vertexaisearch.cloud.google.com, which is a redirector rather than a real source. If you count raw citation URLs without resolving those, you will conclude Google cites Google, which is not what is happening.

Finding 2: we finished last

Brands named across the same 48 answers, by the share of answers that named them:

Zero. Not a low score, an absence. In 48 answers across six engines, no engine named us once.

We could have quietly fixed that before publishing anything. We are publishing it because the zero is the clearest demonstration of the thing we keep telling clients: a good product that no third party has written about is invisible to an AI assistant, and you will not notice, because nobody tells you when you were left out of an answer.

Finding 3: the mechanism, checked by hand

We took the 24 list posts the engines cited most and fetched every one of them, looking for each brand by name.

Peec AI appeared in 20 of the 24. SearchVis appeared in none of them.

That is the whole explanation. The engines were not making a judgment about which product is better. They were reading a set of list posts and repeating the names they found there. Our absence from the answers is not a verdict on the software. It is an accurate report of our absence from the sources.

What this means if you are trying to get recommended

Four things follow directly from the data.

Your website is not the lever you think it is. It has to be good, accurate and easy to read, or nothing else works. But no amount of polish on your own pages puts you in an answer that was assembled from somebody else's list post.

Find the specific pages the engines actually read. Not "get more backlinks". In this category, fewer than fifty pages carry nearly all the weight, and the set is knowable. You can only work the list once you know what is on it.

Reddit and YouTube are not optional. The most cited domain in our category is a forum. The second is a video platform. If you have decided both are beneath your brand, you have decided to be absent from the top two sources in your market.

Being included beats being ranked first. Across our answers, brands that appeared anywhere in a cited list post got named far more often than brands that appeared in none. The gap between position 1 and position 6 inside a list post is real but small. The gap between being in the list and not being in it is everything.

What we are doing about our zero

Publishing this, first. Then working the list: the pages our own tool identified, in the order it ranked them, and reporting back with the numbers whether they move or not.

If you want the same read on your own category, that is exactly what SearchVis does. It asks your buying questions on all six engines, shows you which competitors get named instead of you, and hands you the specific sources you are missing from. The free tier costs nothing and takes a website address.

We will re-run this exact study in ninety days and publish the new numbers, including the ones that make us look bad.