Who AI recommends for AI website chatbots, and which sources it cites
We asked Perplexity and Gemini 10 questions a business shopping for one would ask, three times each (60 answers, October 1, 2026). Tidio was named most: in 15 of 20 question checks. 44 AI chatbot tools were named on at least one check. The most-cited site was tidio.com, in 34 of 60 answers. Asked each question once (October 1, 2026), ChatGPT named Tidio most (in 7 of 10 answers); that single run is counted separately from the numbers above.
By Matt Turley60 answers761 citations10 questionsPerplexity and GeminiChatGPT: 1 run per question, shown separately
A check is one question on one AI tool, asked three times. An AI chatbot tool counts on a check when it was named in at least 2 of the three answers. Perplexity and Gemini together make 20 checks.
AI chatbot tools named by majority, October 2026 (top 15 of 44)
#
Product
Checks named (of 20)
Perplexity (of 10)
Gemini (of 10)
Answers naming it (of 60)
1
Tidio
15
5
10
43
2
Chatbase
9
4
5
29
3
Boei
6
3
3
18
4
Botpress
6
3
3
18
5
SiteGPT
5
3
2
15
6
CustomGPT
5
2
3
12
7
ChatBot.com
4
1
3
13
8
Botsonic
3
1
2
10
9
Freshchat
3
1
2
8
10
Voiceflow
3
1
2
8
11
SiteSpeakAI
3
2
1
7
12
Ada
3
0
3
8
13
Crisp
3
0
3
7
14
My AskAI
2
1
1
7
15
AI Engine
2
1
1
6
Small sample, read it as a snapshot: 10 questions, two AI tools, three runs each, on October 1, 2026. A lead of one or two checks is within the run-to-run noise. This says nothing about the quality of anyone's work, and an AI chatbot tool missing here was simply not named in most runs of these answers.
Table 2 · By AI tool
Perplexity and Gemini only partly agree
18 of the 44 AI chatbot tools named were named by both tools. The rest were named by one tool only.
Each AI tool on the same 10 questions
AI tool
Answers
Named by majority
Named most
Citations per answer
Perplexity sonar
30
30
Tidio (5 of 10 questions, 50.0%)
14.1
Gemini gemini-2.5-flash
30
32
Tidio (10 of 10 questions, 100.0%)
11.2
ChatGPT gpt-4o
Asked once per question, so not comparable to the majority counts here. See the ChatGPT section.
Company websites Directories and review sites Best-of lists and comparisons Reddit, forums, social and video News, docs and reference
Table 3 · Top cited sites
The 15 sites cited most
Counted as answers citing the site at least once, out of 60. These are the pages to be on.
tidio.comCompany websites34 56.7%
eesel.aiBest-of lists and comparisons25 41.7%
chatbot.comCompany websites24 40.0%
sitegpt.aiBest-of lists and comparisons23 38.3%
boei.helpBest-of lists and comparisons22 36.7%
chatbase.coCompany websites22 36.7%
lindy.aiBest-of lists and comparisons18 30.0%
blog.hubspot.comCompany websites13 21.7%
g2.comDirectories and review sites12 20.0%
zendesk.comCompany websites12 20.0%
botpress.comCompany websites11 18.3%
jotform.comCompany websites11 18.3%
yourgpt.aiBest-of lists and comparisons11 18.3%
denser.aiCompany websites10 16.7%
freshworks.comCompany websites10 16.7%
Answers citing each site, by AI tool
Site
Type
Perplexity (of 30)
Gemini (of 30)
tidio.com
Company websites
24
10
eesel.ai
Best-of lists and comparisons
12
13
chatbot.com
Company websites
15
9
sitegpt.ai
Best-of lists and comparisons
14
9
boei.help
Best-of lists and comparisons
12
10
chatbase.co
Company websites
18
4
lindy.ai
Best-of lists and comparisons
7
11
blog.hubspot.com
Company websites
12
1
g2.com
Directories and review sites
12
0
zendesk.com
Company websites
7
5
botpress.com
Company websites
6
5
jotform.com
Company websites
9
2
yourgpt.ai
Best-of lists and comparisons
9
2
denser.ai
Company websites
9
1
freshworks.com
Company websites
6
4
Table 4 · ChatGPT, one run per question, collected October 1, 2026
What ChatGPT named, asked once per question
The same 10 questions went to ChatGPT one time each, not three. So a name counts here when it is written in that one answer. These counts are not in Tables 1 to 3, which stay the Perplexity and Gemini majority.
AI chatbot tools named by ChatGPT, October 1, 2026 (top 15 of 34)
#
Product
ChatGPT answers naming it (of 10)
Share of answers
1
Tidio
7
70.0%
2
Chatbase
5
50.0%
3
Botpress
3
30.0%
4
Freshchat
3
30.0%
5
Ada
2
20.0%
6
Asks
2
20.0%
7
Askvo
2
20.0%
8
Boei
2
20.0%
9
Chatling
2
20.0%
10
Crisp
2
20.0%
11
AI Engine
1
10.0%
12
Animam.ai
1
10.0%
13
Asyntai
1
10.0%
14
BotPenguin
1
10.0%
15
Botsify
1
10.0%
ChatGPT65 citations · 6.5 per answer
42%48%
Company websites Directories and review sites Best-of lists and comparisons Reddit, forums, social and video News, docs and reference
Sites ChatGPT cited most, as answers citing the site at least once (of 10)
asks.appBest-of lists and comparisons4 40.0%
aiideasdb.comCompany websites2 20.0%
askvo.ioCompany websites2 20.0%
g2.comDirectories and review sites2 20.0%
accelrio.comBest-of lists and comparisons1 10.0%
agentforge.solutionsBest-of lists and comparisons1 10.0%
aibotkit.ioCompany websites1 10.0%
animam.aiBest-of lists and comparisons1 10.0%
asyntai.comBest-of lists and comparisons1 10.0%
bestchatbot.ioNews, docs and reference1 10.0%
One run per question is a thin sample: the names in an AI answer move from run to run, so a business named once here could be missing on the next ask. Read it as one look at ChatGPT on October 1, 2026, next to the steadier three-run numbers above.
What to do with it
If you sell AI chatbot tools
Your buyers ask these AI tools what to use. In software, the answers lean on best-of lists and alternatives pages much more than on directories, so the lists in Table 3 matter more than any single review site.
Start with the lists the AI cites most. Find who publishes them and whether you belong there. Then make sure your own site and docs say plainly who you are for, what you replace and what it costs, so the AI can quote you when it compares tools.
Do not judge from one answer. The names move from run to run, which is why every check here is three runs and a name only counts when it shows up in most of them.
See your own product
Are you one of the AI chatbot tools AI names?
This page covers the whole category. The AI Answer Check runs the same test on your own product: the questions your buyers ask, three runs each on ChatGPT, Perplexity and Gemini, whether you are named, who is named instead, and the three fixes worth the most. I read every report before it goes out.
best AI chatbot builder for a small business website
best AI chatbot trained on my own website content
best Chatbase alternative
best Tidio alternative for an AI support chatbot
best AI chatbot for a Shopify store
best AI chatbot plugin for a WordPress website
best AI support chatbot for a SaaS help center and docs
cheapest AI chatbot for a small business website
best white label AI chatbot platform for agencies
AI tools
Perplexity: Perplexity API, sonar model, live web search; cited URLs from the citations field. 30 answers.
Gemini: Google Vertex AI, gemini-2.5-flash with Google Search grounding; cited URLs from grounding chunks. 30 answers.
ChatGPT (separate, one run per question): OpenAI Responses API, gpt-4o with the web_search tool forced on; cited URLs from the answer annotations. 10 answers on October 1, 2026. Shown only in Table 4.
These are the APIs, not the consumer apps. Nobody was signed in and there was no system prompt: the question was the whole input. Each question went to each tool three times on October 1, 2026.
Who counts as named
A list of the products the answers recommend was built first by reading every answer. Then each answer was checked for each name in its text, with links and citation chips removed, so a cited link alone never counts. Spelling variants count as one name: curly or straight apostrophes, hyphen variants, and “&” or “and”. A name counts on a question and tool when it appears in at least 2 of the three runs. Directories, review sites and platforms are never ranked. ChatGPT was asked each question only once, so in Table 4 a name counts when it is in that one answer, with the same spelling rules. Those counts are never added to the majority numbers.
Sources
A citation is a URL the API itself returns as a source (Perplexity's citations list, Gemini's grounding sources resolved from Google's redirect links, and for ChatGPT the URLs in the answer annotations). Each site counts once per answer. Source types come from a written rulebook, not a person reading every page, so some pages land in the wrong group, mostly between company sites, blogs and lists.