Microsoft Power Apps, Retool, ToolJet and Zite each landed in all six answers when we asked ChatGPT and Gemini who builds the best internal tools, a clean 6 of 6 (100%) for four products at once. No single winner ran away with it. There is a four-way tie at the top, and that tie holds no matter how the question is phrased.
Softr was the clear runner-up at 5 of 6 answers (83%), missing only one framing. Behind it, Zapier and Zoho Creator tied at 4 of 6 (67%), and Budibase and Glide tied at 3 of 6 (50%). Everything below that was a long tail: Appsmith at 2 of 6 (33%), then DronaHQ, Google AppSheet and Stackby at 1 of 6 (17%) each. Thirteen products were named in total across the six answers.
We put six buyer questions to ChatGPT and Gemini, one live web search each: the bare "best internal tool builders," then "for small teams," "what should I use," "best free," "most affordable," and "recommendations." Recording every product and every source domain in all six answers is the whole exercise. The numbers below are counts out of those six answers, nothing modeled or weighted.
How the shortlist changed from "best" to "free" to "affordable"
The four constants, Microsoft Power Apps, Retool, ToolJet and Zite, appeared in every one of the six sub-questions, 6 of 6. If you only read the top of each answer, the category looks settled. The real movement is underneath, and it is large.
The bare "best" question returned eight names including Softr, Zapier and Glide. Switch the wording to "best free" and those three vanish, replaced by the open-source and self-hosted set: Appsmith (in 2 of 6 answers overall), Budibase (3 of 6) and DronaHQ. DronaHQ showed up in that "best free" answer and nowhere else, 1 of 6. So the word "free" alone swapped out three commercial tools and swapped in three self-hostable ones.
"Most affordable" also returned eight names, and it sat between the two. It kept Softr and Zapier from the "best" list, kept Zoho Creator (4 of 6 overall), and added Budibase (3 of 6), but it did not pull in Appsmith or DronaHQ. In this data "affordable" reads closer to "best" than "free" does: paid tools with cheap tiers count as affordable, while "free" specifically surfaces the open-source names.
Two other products each lived in a single sub-question. Stackby appeared only in "best internal tool builders for small teams," 1 of 6. Google AppSheet appeared only in "what internal tool builders should I use," 1 of 6. Together with DronaHQ, that is three products whose entire presence in the category rests on one phrasing of one question. Ask the category a slightly different way and they are gone.
Softr's one absence is also telling: it scored 5 of 6, and the single answer it missed was "best free." Zapier's two misses (it scored 4 of 6) were "small teams" and "best free." The free and small-team framings are where the shortlist reshuffles hardest.
The source map: who fed the answers
Reddit fed all six answers, 6 of 6, the only community forum that appeared. No Quora, no other forum showed up. The rest of the sources were vendor blogs and roundup listicles: stackby.com (6 of 6), zite.com (6 of 6), reflex.dev (5 of 6), softr.io and zapier.com (4 of 6 each), blog.tooljet.com and fillout.com (3 of 6 each), weweb.io (2 of 6), and baserow.io, blaze.tech, viasocket.com and youtube.com at 1 of 6 each.
The key insight is what happens when a vendor's own domain is in that list. zite.com fed 6 of 6 answers and Zite was named in 6 of 6. softr.io fed 4 answers and Softr was named in 5. zapier.com fed 4 and Zapier was named in 4. blog.tooljet.com fed 3 and ToolJet was named in 6. In every one of those cases the company publishing the page is also the product the assistant recommends. Your own domain in the source set is not a vanity metric here, it is a naming channel.
There is a sharp counter-example in the same data. stackby.com fed 6 of 6 answers, tied for the most of any domain, yet the product Stackby was named in only 1 of 6. Stackby's listicle is being cited as a neutral roundup, not as the pick. Feeding the answer and being the answer are two different outcomes, and a roundup can do the first without ever earning the second.
One absence matters as much as any presence: G2 and Capterra did not appear in a single answer, 0 of 6. The classic review directories fed nothing here. Community threads and vendor roundups did all the work.
What a vendor in this category would do to get named
Presence in these exact sources, not market share, decided who got named. Microsoft Power Apps is enormous and it scored 6 of 6, but so did Zite, which is far smaller, because zite.com and the Reddit threads both carried it. The assistant reads a source list, then names what that list contains.
So the playbook is concrete. Be in the Reddit threads that fed 6 of 6 answers. Get into the roundups that carried real weight: reflex.dev (5 of 6), fillout.com (3 of 6), weweb.io (2 of 6). And publish your own comparison page, because every vendor whose domain fed an answer, zite.com, softr.io, zapier.com and blog.tooljet.com, also saw its product named. Chasing a G2 or Capterra profile would have returned nothing in this measurement, since neither appeared in any of the six answers.
The full list, counted
| Product | Named in | Share |
|---|---|---|
| Microsoft Power Apps | 6 of 6 | 100% |
| Retool | 6 of 6 | 100% |
| ToolJet | 6 of 6 | 100% |
| Zite | 6 of 6 | 100% |
| Softr | 5 of 6 | 83% |
| Zapier | 4 of 6 | 67% |
| Zoho Creator | 4 of 6 | 67% |
| Budibase | 3 of 6 | 50% |
| Glide | 3 of 6 | 50% |
Glotier does not sell an internal tool builder, so we are correctly absent from all six of these answers, and we would not want it any other way: naming a tool the data did not surface is exactly the fabrication this series exists to avoid. But the measurement on this page is the thing we run for a customer, pointed at their own category and their own product. The check is free, needs no account, and takes about a minute.