Statsig was the one name that carried this category: it appeared in 5 of the 6 answers ChatGPT, Gemini and Perplexity gave when we asked, on 2026-08-03, who makes the best A/B testing tools. No other product matched it. AB Tasty, Firebase A/B Testing and VWO tied for second, each named in 4 of the 6 answers. We put six buyer questions to the three engines, ran one live web search per answer, and wrote down every product named and every source domain behind it. Everything below comes from that single run.
Statsig's 5 of 6 is a lead, not a landslide. Being named in 83% of answers still left it absent from one question, and that question was "most affordable a/b testing tools," where Statsig did not appear at all. The three runners-up (AB Tasty, Firebase A/B Testing and VWO, each 4 of 6, or 67%) each missed two of the six questions. In a six-question sample, the distance from first place to fourth is a single answer, so no vendor here owns the category the way a market-share chart might imply.
Below the top four sits a dense pack at 3 of 6, or 50% each: Crazy Egg, GrowthBook, Optimizely, PostHog and Qualaroo. Five products tied on exactly half the answers. Then a cliff. Google Optimize and LaunchDarkly landed at 2 of 6 (33%), and a long tail of twelve products was named just once (17%), including ABRouter, Adobe Target, Mida, Microsoft Clarity, Unbounce and Zoho PageSense. Twelve of the 23 distinct products we recorded, more than half, were named in only a single answer.
How the shortlist changed across the six questions
The shortlist was not one list. It reorganized itself depending on which word we added to the question, and that is the finding that matters most for anyone trying to get named. The bare "best a/b testing tools" question returned an enterprise-leaning set of 8: Optimizely, Google Optimize, Unbounce, AB Tasty, Crazy Egg, Statsig, Firebase A/B Testing and LaunchDarkly, with Unbounce appearing in this question and nowhere else across the six.
"Best free a/b testing tools" pulled the list somewhere else entirely. That one question surfaced ABRouter and Mida, two names that appear in none of the other five answers, alongside the open-source pair PostHog and GrowthBook. The word "free" was enough to swap in tools a generic "best" query never mentioned, out of 8 named for that question.
"Most affordable a/b testing tools" produced its own one-off: Zoho PageSense, named only in this question. It is also the single question Statsig missed, so the category leader was invisible exactly where price was the filter. GrowthBook and PostHog carried over from the free question into this one, which suggests the models treat "free" and "affordable" as neighboring, open-source-friendly buckets.
"Best a/b testing tools for small teams" had three exclusives of its own out of 8 named: Microsoft Clarity, PageSense and VWO Testing, each appearing only in the small-teams answer. And "a/b testing tools recommendations" was the only place Adobe Target and Google Analytics 4 turned up. Across six questions, five of them surfaced at least one product that lived in that question alone.
The practical read is the split down the middle. GrowthBook and PostHog were each named 3 of 6 times, but only in the small-teams, free and affordable questions, never in the bare "best" or "recommendations" answers. Optimizely was the mirror image, named 3 of 6 but only in "best," "what should I use" and "recommendations," never once in free or affordable. The same category divides into a paid-incumbent list and an open-source list depending on a single adjective.
The source map: who fed the answers
Two community forums fed every single answer. Reddit appeared in 6 of 6 answers and Quora in 6 of 6, making them the only two sources with perfect coverage across the run. Whatever else the engines read, they read Reddit and Quora threads about A/B testing every time.
After the forums came an analyst and two vendor blogs. Gartner fed 5 of 6 answers. So did qualaroo.com and uxcam.com, two vendor-owned domains running "best A/B testing tools" roundup posts. convert.com fed 3 of 6 and coursera.org 2 of 6. Everything else was a single appearance: blog.hubspot.com, conversion-rate-experts.com, crazyegg.com, medium.com, mida.so, optimizely.com, sitepoint.com, unbounce.com and vwo.com each showed up in exactly 1 of 6 answers, for 17 distinct domains in total.
The review directories were mostly missing. Capterra did not appear in any of the six answers. G2 appeared once, and only through its blog subdomain learn.g2.com, in 1 of 6. For a software category, that is worth pausing on: the buying guides the engines leaned on were Reddit, Quora and Gartner, not the directory listings a vendor usually pours effort into.
Here is the insight we would circle for anyone selling in this category. qualaroo.com fed 5 of 6 answers, and Qualaroo the product was named in 3 of 6. The vendor that published one of the most-cited roundups landed itself in half the answers. But it is not automatic. uxcam.com also fed 5 of 6 answers, yet UXCam the product was named in only 1 of 6. Publishing the list the engines read gets your domain cited. It does not force them to repeat your own name.
The opposite pattern is the healthier one. VWO's own domain, vwo.com, was cited just 1 of 6 times, yet VWO the product was named 4 of 6. optimizely.com and crazyegg.com were each cited once, while Optimizely and Crazy Egg were each named 3 of 6. Those products are named because Reddit, Quora, Gartner and other people's roundups vouch for them, not because their own site seeded the answer.
What a vendor here would do to get named
If we sold an A/B testing tool, we would not chase market share to get named. We would chase these 17 domains. The answer set was decided by presence in a specific, countable source list: two forums in 6 of 6, Gartner in 5 of 6, two roundup blogs in 5 of 6, and a tail of single-appearance pages. Get discussed in the Reddit and Quora threads the engines actually pulled, get into the Gartner listing, and get placed in the roundups that rank, and you are in the pool. Stay out of them and you are invisible, however large you are.
The single-mention tail proves it. ABRouter and Mida are not big companies, but each was named 1 of 6 because it sat in the one free-tools source that answered the "free" question. Presence beat size. A product's odds in these answers track the domains it appears on, not the number of customers it has.
The full list, counted
| Product | Named in | Share |
|---|---|---|
| Statsig | 5 of 6 | 83% |
| AB Tasty | 4 of 6 | 67% |
| Firebase A/B Testing | 4 of 6 | 67% |
| VWO | 4 of 6 | 67% |
| Crazy Egg | 3 of 6 | 50% |
| GrowthBook | 3 of 6 | 50% |
| Optimizely | 3 of 6 | 50% |
| PostHog | 3 of 6 | 50% |
| Qualaroo | 3 of 6 | 50% |
Where this leaves us
Glotier is not on this list, and that is correct. We do not make an A/B testing tool, so ChatGPT, Gemini and Perplexity should not name us when a buyer asks for one, and we would not want them to. But the run you just read is exactly the thing we do. We take one category, put the six buyer questions to the three engines, and hand back the named products, the counts and the source domains behind every answer.
If you sell one of the tools above, or wish you were on the list, the same measurement runs on your own category in about a minute. It is free and there is no account. You would see your own "named in X of 6," the questions where you are missing, and the exact domains deciding it, which is the first honest picture of whether the AI answers are working for you or for someone else.