Asktopus case study: named in 0 of 138 AI answers

Why track our own product?

Asktopus is an AI chatbot for websites, made by Uneven Bits, the company behind sirefs. We run sirefs on our own products before anyone else's, because we know them well enough to check every number by hand. This is the first run on Asktopus, reported in full: the prompts, the engines, every denominator and what the data recommends.

We make both products, so we have an interest in the result. That's why the answers behind every number are public in the live demo.

How was the study set up?

Which questions did we ask?

The ones a buyer asks before choosing a chatbot: which is best, what the alternatives to well-known tools are, what it costs, and whether a tool handles a specific need. Topics run from general best-of questions to ecommerce, pricing, answering from your own content, GDPR and handing a conversation to a human.

Twelve of the 16 prompts name no brand. Four ask for alternatives to a named competitor (Chatbase, Tidio, Intercom Fin and Karla). None names Asktopus. The full list is at the end of this page.

Which engines, and how were the answers collected?

Microsoft Copilot isn't collected yet, so it isn't in this study. How sirefs collects answers explains each method.

Why 138 answers and not 112 (16 prompts on 7 engines)? On that first day, 26 checks ran twice, a few minutes apart, and the dashboard keeps both answers. Five answers have no text: 2 Google searches showed no AI Overview, and 3 AI Mode responses came back blank. They count as answers without a mention, as they do on the dashboard. Leaving them out gives Asktopus 0 of 133 (0% to 2.8%) and changes no conclusion.

How did we count?

A brand is mentioned when its name or an alias appears in the answer, found either by name matching or by a language model reading the answer. A brand named in the question doesn't count for that question, which is why some denominators are below 138. Every rate comes with a 95% Wilson interval: the range the true rate very likely falls in. Reading the numbers explains the method.

We also read the answers behind the counts, and found a bug. In three Grok answers, the language model listed tracked brands the answers never name, and the dashboard counted them: 14 mentions in all, including a single Asktopus mention. We fixed it on 8 October 2026: a brand the language model lists now counts only when its name is in the answer's text. The numbers on this page, and in the demo, are the corrected ones.

How often did the engines name Asktopus?

Never. Asktopus was named in 0 of 138 answers (95% interval 0% to 2.7%). asktopus.com was cited in 0 of 138, although 128 answers cited at least one page.

Read the engine rows with care: 18 to 22 answers can't pin a rate down. Zero of 18 is consistent with a true rate anywhere up to 18%. The pooled figure is the firm one: across 138 answers, the true rate is very likely under 3%.

The markets tell the same story: 0 of 99 US English answers (0% to 3.7%), 0 of 28 Danish (0% to 12%) and 0 of 11 Spanish (0% to 26%). Asktopus is made in Copenhagen, and the Danish answers didn't name it either.

Rank doesn't apply: a brand that's never named has no position.

Who did the engines name instead?

At least one of the nine competitors we track was named in 124 of the 133 answers with text (93%, 88% to 96%).

Tidio came first in 34 list-style answers, more than any other brand, and the engines recommended it, not just named it, in 81 of the 128 answers the language model read (63%, 55% to 71%). Positions move a lot from one answer to the next, so read them as a tendency.

The engines also named brands we don't track. SiteGPT was named in 29 of the 136 answers the model read, by six engines, and recommended in 23: as often as Gorgias and Crisp. Boei (16 answers), CustomGPT.ai (14), eesel AI (12), Shopify Inbox (12), Freshdesk (12) and Drift (12) followed.

Which sources did the engines cite?

128 of the 138 answers cited at least one page, and none was on asktopus.com. All ten of the most cited domains belong to companies that sell chatbots or support software: five we track, and five we don't, cited mostly for blog posts that compare tools.

No answer named Asktopus, so every one of these sources appeared next to competitors and never next to us. The single most cited page was a Boei post on the best AI chatbots of 2026, in 13 answers. eesel.ai was the only domain all seven engines cited, most often for its Chatbase alternatives post (5 answers).

Beyond the vendors, YouTube (11 answers) and Reddit (10) were cited only by Google AI Overviews and AI Mode, and G2 by four engines (9). In Danish, two local sites were cited in 9 answers each: businesswith.dk, mostly its Karla alternatives page, and chatassist.dk, for its pricing and Karla alternative pages.

What did the GEO audit find?

The audit reads a site the way an AI crawler does. asktopus.com scored 91 of 100 on 6 October 2026.

The site publishes llms.txt, a crawlable pricing page, FAQ and product markup, and 18 comparison or alternatives pages. It lost points for having no sameAs links to its profiles and no Wikidata entry. None of its pages was cited in this run. The audit says whether engines can read and trust a site, not whether they'll choose it, and here they didn't.

What did the answers say?

Three examples from 6 October 2026, quoted as collected.

For most small-business websites, I’d pick Tidio. It’s a good balance of AI, live chat, ease of setup, and price.

ChatGPT (ui), United States, English. Prompt: "What is the best AI chatbot for a small business website?"

The best AI chatbot for your website depends on your technical skill and main goal, with ChatBot offering the easiest all-in-one setup for general customer service.

Google AI Overviews (serp), United States, English. Prompt: "best ai chatbot for website". The only source for that sentence was chatbot.com, ChatBot's own site.

Hvis du ikke allerede har et kundeservicesystem, ville jeg starte med Tidio/Lyro.

"If you don't already have a customer service system, I'd start with Tidio/Lyro." ChatGPT (ui), Denmark, Danish. Prompt: "Hvad er den bedste AI chatbot til en hjemmeside?" Its comparison table listed five tools, all on our competitor list: Intercom Fin, Tidio, Chatbase, Zendesk AI and Botpress.

What does it mean?

A well-built site wasn't enough. In this category, the engines answered from a short list of established brands and from comparison posts that vendors write about each other. Asktopus has comparison pages of its own; the engines cited other people's.

For anyone deciding whether this kind of tracking is worth it: none of this shows up in web analytics. A buyer who asks ChatGPT and gets five names that aren't yours never reaches your site. You only see the gap by asking the questions, on every engine, with enough answers to trust the number.

What does the data recommend?

These are the dashboard's next steps for the project on 8 October 2026, worked out from the answers above. They're recommendations. Nothing on this page shows whether they work: that takes later runs and a significance test.

1. Answer three questions where every answer names a competitor. "¿Cuál es el mejor chatbot con IA para una página web?" (all 11 answers; Tidio in 10), "Hvilke alternativer er der til Karla chatbot?" (all 11; Intercom in 10) and "What are the best Chatbase alternatives?" (all 10; Intercom and Botpress in 10 each). The answers show the format the engines quote and the pages they cite.
2. Get mentioned on the sites the engines cite. eesel.ai was cited in 19 answers and chatbot.com in 18, none of which named Asktopus. Both belong to other chatbot vendors, so the lesson is as much the format as the placement: they're cited for best-of and alternatives posts.
3. Track the brands the engines name. SiteGPT (29 answers, recommended in 23) and Boei (16 answers, recommended in 13) aren't on the competitor list yet.
4. Take part where the conversation is cited. Google AI Mode cited the Reddit thread "Top 3 AI Chatbots for E-commerce SMBs in 2026" in 2 answers. If you can genuinely help there, reply as yourself.

The audit adds one fix: list the brand's real profiles in the sameAs field of the site's Organization markup.

What are the limits of this study?

• One run, one day. All 138 answers were collected on 6 October 2026, within 16 minutes. It's a snapshot, not a trend. The project runs weekly; later runs aren't on this page.
• Sample size. 138 answers give a firm result for Asktopus overall, but each engine gave 18 to 22 answers and the Spanish market 11, so per-engine and per-market rates have wide intervals.
• Repeat answers. 26 checks ran twice, a few minutes apart. Two answers minutes apart are less independent than two answers a week apart.
• Methods differ. ChatGPT and Gemini are logged-out app answers, and Google's two are what its results page showed. Perplexity, Claude and Grok answered through their APIs (Sonar, Claude Haiku 4.5 and Grok 4.3), which often cite different pages and name different brands than the apps do. Microsoft Copilot isn't covered.
• One product, one niche. Website chatbots is a category with large, well-known players, and we wrote the 16 prompts. Another product, category or prompt set could look very different.
• What counts as a mention. Any naming counts, including passing ones, such as Zendesk named as a place to send escalations. Recommendations and positions come from a language model's reading and are noisier than mention counts.
• Our own product. We make Asktopus and sirefs. The answers are public, so you don't have to take our word for it.

How can you check these numbers?

The live demo is this project: the 16 prompts, every answer with its engine, method, market and date, the competitors, cited sources, next steps and audit. The answers in this study are dated 6 October 2026. The project keeps running weekly, so the demo's totals will move as new answers arrive.

To see where your own brand stands, the free check audits your site and asks three buyer questions on ChatGPT, Google AI Overviews and Perplexity. No sign-up.

All 16 prompts