Does AI know Llamat? Our baseline: 0 of 66 answers

This post is about our own visibility. The numbers come from the same measurement as our monthly dataset Which AI visibility tools do ChatGPT and Gemini recommend?. We did not change the phrases, the runs or the counting rules for this post, and we will publish later rounds whether the numbers move or not.

The baseline: named in 0 of 66 answers

The 11 phrases are searches people really type about tools like ours, taken word for word from US search suggestions on the day: “best ai visibility tools”, “best ai visibility tracking tools”, “best ai visibility platforms”, “best ai search visibility tools”, “ai visibility checker free”, “ai search visibility tool free”, “best aeo tool”, “best llm seo tools”, “best generative engine optimization tools”, “best ai brand monitoring tools” and “best chatgpt visibility tracker”. We asked each one three times on ChatGPT and three times on Gemini, every time in a new logged-out session. That gave 66 answers, and all 66 were successful.

ChatGPT named Llamat in 0 of its 33 answers, and Gemini in 0 of 33. Neither engine cited llamat.com as a source. Two days earlier, on 1 October, logged-out ChatGPT had not named it in any of 12 answers to four of the same phrases either. In the terms of our own report, every rate is 0%: mention, recommendation and citation. The AI Visibility score is 0, and there is no average position, because position is only counted in answers that name the brand.

The category itself was crowded. The answers named 59 different tools: 29 on ChatGPT, 44 on Gemini and 14 on both. An average ChatGPT answer named 6.55 tools and an average Gemini answer 7.12. Five names dominated: Semrush appeared in 57 of the 66 answers, Otterly.AI in 54, Ahrefs in 53, Profound in 48 and Peec AI in 47. After SE Ranking / SE Visible (23) and Scrunch (20), every other tool appeared in 7 answers or fewer.

Why a 12-day-old site is expected to be absent

We can see the answers, not how the engines chose them, so what follows is our reading of the data, not a proven cause. Four things point the same way.

• No history. The public registration record shows that llamat.com was registered on 21 September 2026, twelve days before the measurement. Whatever the models behind these engines learned before that date cannot include anything published on it.
• The answers came with sources attached. Every one of the 66 answers cited at least one source: 5.21 domains per ChatGPT answer on average and 6.88 per Gemini answer. The names in the answers came with pages behind them.
• Those pages do not mention us. The most-cited pages were “best AI visibility tools” roundups on other sites. Zapier's list was cited in 19 of 33 Gemini answers, a buyer guide on dataforaisearch.com in 14 of 33 ChatGPT answers, Position Digital's reviews in 13 Gemini answers and a Marqeable comparison in 7 ChatGPT answers. Gemini also cited Reddit in 14 of its 33 answers, through two threads about the best AI visibility tools. We read those four roundups and SE Visible's own list of the best AI visibility tools on 3 October 2026, and Llamat is in none of them.
• Few links, little time. A site this young has had little time to earn links, reviews or a place in anyone's list. Pages of our own on the topic, such as our comparison of AI visibility tools, only help once an engine picks them as a source, and on 3 October no answer cited any page on llamat.com.

Age alone is not the whole story, though. Zapier's list is dated 14 September 2026, Position Digital's was updated on 16 September, and Data for AI Search says it verified its prices on 26 September. The engines were citing pages written or updated in the three weeks before the measurement. Vendors' own sites were sources too: ahrefs.com in 10 ChatGPT answers and 5 Gemini answers, and visible.seranking.com in 13 Gemini answers. So a recent page can be a source, at least on an established site. What Llamat lacks is a place on the pages the engines already pick for these searches, or pages of its own that they pick.

The top of the list is settled, the tail is not

Between 1 and 3 October, on the four phrases both measurements share, logged-out ChatGPT named the same five tools most often. Ahrefs was in 12 of 12 answers (10 two days earlier), Semrush in 11 (10), and Otterly.AI, Peec AI and Profound in 9 each, as before. Further down, Promptwatch fell from 6 answers to 2 and AthenaHQ from 5 to 3. Both days named 15 different tools, and 13 of the 17 tools named on either day were named on both.

Within a single day the tail moves too. Counting every tool named for a phrase, ChatGPT named 64% of them in all three runs and 21% in only one. On Gemini it was 46% in all three and 34% in only one. A newcomer could first show up as a single mention in one run, and that is also exactly what noise looks like. This is why one mention will not count as progress on its own.

What we will do

The route we want is the slow one: publish things worth citing, and let other people decide whether to cite them.

• Publish our own data. Our monthly dataset Which AI visibility tools do ChatGPT and Gemini recommend? includes every answer as a CSV file, and our study of which websites block AI crawlers covers the robots.txt files of the 10,000 most popular websites. Numbers nobody else has are the main thing a new site can offer.
• Publish the method. Our methodology explains how every number is collected and counted, including the parts that look worse for us: three runs, asked once, and ChatGPT asked through Czech IP addresses for every market.
• Keep our comparisons honest. We read vendor facts from the vendors' own pages and date them, and every page that compares Llamat with other tools says that Llamat is our product.
• Tell the people who write about AI search when we publish something new, and let them judge whether it deserves a mention.

What we will not do

There are shortcuts, and the sources above make them tempting, since Reddit threads and “best tools” lists are exactly what the engines cited. We will not take them:

• post about Llamat on Reddit, forums or review sites under made-up names, or without saying who we are;
• write or buy reviews, or pay for undisclosed places in other people's “best tools” lists;
• ask the engines about Llamat over and over in the hope that they pick it up;
• change the 11 phrases, add our name to them, or swap in phrases because they suit us.

How and when we will measure again

Every quarter, with nothing changed. We will use the same 11 phrases word for word, each asked three times on ChatGPT and three times on Gemini in new logged-out sessions. Both engines are asked through our own proxy farm with Czech exit IPs; Gemini also gets the United States as its country, which did not change the location it saw in a test on 4 October 2026. We use Czech exits for ChatGPT because that is where our own proxy farm runs, and ChatGPT has no country setting that would let us pick a market instead; our methodology lists what that can change. On 4 October we checked all 33 ChatGPT answers in this baseline for a local slant: all are in English, none mentions the Czech Republic and none cites a .cz site. That does not prove a US visitor would get the same answers, and we have not measured that. The counting stays the same too: Llamat counts as named when its name appears as a whole word in an answer (links removed first), and as cited when llamat.com is among the answer's sources. Mentions and citations are matched by rule, the same way our report matches them.

The same phrases are planned to run monthly for the dataset, and Llamat's numbers are part of every release. This post will get a dated update each quarter, and the next one is planned for January 2027. If a phrase stops appearing in search suggestions, we will keep asking it so the rounds stay comparable, and say so here. If an engine stops answering logged-out visitors, as Perplexity did in our tests (0 of 12 answers on 1 October 2026), we will report the gap instead of quietly switching to a different method.

What would count as progress

We are writing this down before we see the next result. From weakest to strongest:

• A page that the answers cite for these phrases mentions Llamat. We will read the most-cited pages in each round, as we did on 3 October. This happens upstream of the answers and is not a result in itself.
• Llamat is named in at least one answer. On its own, a single mention is weak evidence: on 3 October, 21% of the tools ChatGPT named for a phrase and 34% of those Gemini named appeared in only one of the three runs.
• Llamat is named in at least two of three runs for the same phrase on the same engine. A Llamat report uses the same majority rule for its ChatGPT and Gemini heatmap tiles.
• llamat.com is cited as a source in any answer.
• Any of the above holds for two quarterly rounds in a row.

Some results would not count. Being named when someone asks about Llamat by name does not count, and this measurement never asks that. Answers from signed-in accounts do not count, and neither do results on phrases picked after we have seen them.

We will not claim to know why something changed, either. Engines change their models and sources without notice, and lists move on their own: Promptwatch dropped from 6 to 2 of 12 ChatGPT answers in two days, and nothing in our data shows why. If Llamat starts to appear, we will report when and where, not that a particular action worked. If it is still 0 of 66 in January, that goes here too.

Limits of this baseline

• One market (en-US), 11 phrases, and three runs per phrase on each engine, collected on one day.
• ChatGPT has no country setting. It answered in English from Czech IP addresses. Gemini was asked with its country set to the United States, but through the same Czech IP addresses (in a test on 4 October 2026 with the country set to the United States, Gemini placed the visitor in Prague, Czechia).
• Logged-out answers only. Signed-in users with Memory or chat history may see different answers.
• These are category searches. They say nothing about whether either engine can describe Llamat when asked about it directly; we did not test that.
• Named is not the same as recommended. We counted every mention of a tool, including one in a closing question. For Llamat the difference does not matter yet.

Run the same test on your own brand

You do not need our tools to repeat this measurement. It takes five steps:

• Fix the phrases before you look at any answer. Use searches people really type about your category, not your brand name; the suggestions a search engine shows as you type are a good place to find them. Keep the same list for every later round.
• Ask each phrase in a fresh logged-out session: a private browser window, not signed in, closed after every answer. Do it three times per phrase on each engine your buyers use, and note the country you asked from, because the IP address can change the answer.
• Save every answer together with its list of sources before you count anything.
• Count by rule. Your brand is named when its name appears as a whole word in the answer text, with links removed first, and cited when your domain is among the sources. Count competitors the same way.
• Read by majority. Treat your brand as present for a phrase on an engine only when it is named in at least two of the three runs, and report shares of answers per engine, not a single screenshot.

Our guide to an AI visibility audit walks through the same procedure by hand, with a spreadsheet template for logging the answers. If you would rather not do it manually, our free AI visibility checker runs it for up to 20 prompts written from real search phrases about your site, three runs each on ChatGPT and Gemini and one search each for Google AI Overviews, and e-mails you the results, typically within 30 minutes after you confirm your address.