DokuWiki was the only tool ChatGPT and Gemini named in all six of six answers when we asked them to recommend wiki software on 2026-08-03. We put six buyer questions to the two assistants, gave each one a single live web search, and wrote down every product they named and every source domain they read. Wiki.js came second, named in 5 of 6 answers. After that the list dropped to a three-way tie at 4 of 6: MediaWiki, Notion and Nuclino.
We run these checks in the open because the result tends to surprise people. Wiki software has plenty of well-funded commercial names, yet the tool that landed in 100% of answers, 6 of 6, was DokuWiki, an open-source project. What follows is just the raw count, phrasing by phrasing, plus the exact pages ChatGPT and Gemini used to get there.
The tools that showed up most
Across all six wiki software questions, five tools were named in four or more of the six answers. DokuWiki led at 6 of 6 (100%), Wiki.js followed at 5 of 6 (83%), and MediaWiki, Notion and Nuclino tied at 4 of 6 (67%) each. Those five names carried the category for ChatGPT and Gemini on 2026-08-03.
Below that top group, four tools were named in exactly 2 of 6 answers (33% each): Atlassian Confluence, Heroic Knowledge Base, Slite and TiddlyWiki. Seven more were named just once, 1 of 6 (17%): BookStack, Confluence, Guru, ProProfs Knowledge Base, Slab, XWiki and Zoho Wiki. Sixteen distinct product names came up in total across the six answers.
The shortlist changed with every phrasing
The six questions were close in meaning, but the shortlist reshuffled with each one, and this is the part worth studying. The bare "best wiki software" query pulled the widest net: 8 names, mixing open-source (MediaWiki, DokuWiki, Wiki.js) with commercial tools (Atlassian Confluence, Notion, Slite, Nuclino) and a WordPress plugin (Heroic Knowledge Base).
Add two words and the list narrows. "Best wiki software for small teams" returned only 4 names: Nuclino, BookStack, Notion and DokuWiki. BookStack appeared in this small-teams answer and nowhere else in the six, its single 1 of 6 showing up entirely on the team framing.
"What wiki software should I use" returned 8 names and was the only place Slab and Guru appeared, each named 1 of 6 on that phrasing alone. It was also the one answer that used the bare label "Confluence", separate from the "Atlassian Confluence" label that the "best" and "best free" answers used for the same product.
"Best free wiki software" swapped the mix toward self-hosted open-source: MediaWiki, DokuWiki, Wiki.js, Heroic Knowledge Base, Notion, Atlassian Confluence, XWiki and TiddlyWiki, 8 names in all. XWiki appeared only here, 1 of 6, surfacing purely on the free framing.
"Most affordable wiki software" collapsed the hardest. It returned just 3 names: DokuWiki, Wiki.js and TiddlyWiki, all free and self-hosted. Every commercial tool that showed up under "best", Notion, Slite, Nuclino and Confluence, disappeared the moment the question mentioned price.
TiddlyWiki tells the fan-out story in a single product. It was named in 2 of 6 answers, and both were the price-conscious ones: "best free" and "most affordable". It never surfaced under "best", "small teams", "what should I use" or "recommendations". Meanwhile "wiki software recommendations" brought its own pair of one-timers, ProProfs Knowledge Base and Zoho Wiki, each named 1 of 6 on that phrasing only. Same category, six wordings, and a product could be top of one answer and invisible in the next.
Where the answers came from
Two domains fed all six answers: reddit.com (6 of 6) and dokuwiki.org (6 of 6). A Reddit thread sat behind every single wiki software answer ChatGPT and Gemini gave on 2026-08-03. Next came herothemes.com and js.wiki, each feeding 5 of 6 answers, then en.wikipedia.org and peoplemanagingpeople.com at 4 of 6 each.
Community Q&A carried real weight in this category. quora.com fed 3 of 6 answers, softwarerecs.stackexchange.com fed 3 of 6, and webapps.stackexchange.com plus discuss.privacyguides.net fed 1 each. Add Reddit's 6 of 6 and forum threads were somewhere in every answer the two assistants produced.
What did not appear is as telling as what did. Not one software review directory showed up in any of the six answers: no G2, no Capterra, nothing of that kind. In this category, on this date, ChatGPT and Gemini built their wiki software answers out of Reddit threads, Wikipedia, a couple of roundups, and the products' own sites, not the star-rating directories many vendors pour budget into.
The vendors whose own pages got them named
Here is the pattern we would circle for any vendor. For several tools, the number of times their own domain was retrieved matched the number of times the product was named. dokuwiki.org fed 6 answers and DokuWiki was named in 6. js.wiki fed 5 and Wiki.js was named in 5. slite.com fed 2 and Slite was named in 2. getguru.com fed 1 and Guru was named in 1. proprofskb.com fed 1 and ProProfs Knowledge Base was named in 1.
Read the other way: the wiki tools ChatGPT and Gemini named most often were the ones whose own documentation and comparison pages the assistants actually pulled. That is the single most useful thing this measurement shows a vendor. The one place the pattern bends is herothemes.com, which fed 5 of 6 answers but as a roundup listing many competitors, so its own Heroic Knowledge Base was named in only 2 of 6. A page that lists everyone gets read; a page that makes your case gets you named.
What we would do if we sold wiki software
If we ran marketing for a wiki tool, we would not start from market share. We would start from this source list, because the source list, not the size of the company, decided who got named. Atlassian Confluence is one of the biggest names in the category, and it was named in only 2 of 6 answers, with the bare "Confluence" label adding 1 more, behind an open-source project sitting at 6 of 6.
Concretely, we would work our way into the exact pages these answers came from: the Reddit threads that fed all 6 answers, the HeroThemes roundup that fed 5, the People Managing People roundup that fed 4, the Quora and softwarerecs.stackexchange threads that fed 3 each, and our own docs and comparison pages, the way dokuwiki.org and js.wiki did. Then we would re-run the same six questions and watch whether our name moved off 0 of 6 and up the list.
The full list, counted
| Product | Named in | Share |
|---|---|---|
| DokuWiki | 6 of 6 | 100% |
| Wiki.js | 5 of 6 | 83% |
| MediaWiki | 4 of 6 | 67% |
| Notion | 4 of 6 | 67% |
| Nuclino | 4 of 6 | 67% |
| Atlassian Confluence | 2 of 6 | 33% |
| Heroic Knowledge Base | 2 of 6 | 33% |
| Slite | 2 of 6 | 33% |
| TiddlyWiki | 2 of 6 | 33% |
| Confluence | 1 of 6 | 17% |
Confluence appears twice in that table on purpose. Our raw count recorded "Atlassian Confluence" in 2 of 6 answers and a bare "Confluence" in 1 of 6, and we are leaving them as two separate rows rather than merging them into a 3 the data never showed. Below this table, also at 1 of 6 each, sit BookStack, Guru, ProProfs Knowledge Base, Slab, XWiki and Zoho Wiki.
Glotier does not sell wiki software, so we are correctly absent from all six of these answers, and we would not want a check that pretended otherwise. But this is exactly the measurement we run for a customer's own category: the six questions, the named products, and the source pages behind each answer. If you sell in a category like this one, the check is free, needs no account, and takes about a minute.