Our criticism of Brave becomes an image result
Yesterday, we documented a limitation in Brave’s AI answer: it described our experiment but listed only 20 field notes. At the time, 24 had already been published. Today, we repeat the same query in web search, image search and “Ask Brave”.
Something stands out beyond a changed results list: the evidence we used to criticise the outdated AI answer now appears in Brave’s own image search.
Our observation becomes an image result
We find the four screenshots from field note 025 in image search: the earlier web search, image search, and the opening and detailed view of the AI answer. They include the table explicitly listing the older total of 20 entries.

We publish a screenshot to make a statement verifiable. The search engine subsequently includes that evidence in its visible image results. Our criticism becomes discoverable content in its own right.
We open one of the new results to check it: the web-search screenshot from 14 October. Its detail view links to our English field note 025 and the original file brave-web-2026-10-14.jpg.

This demonstrates that our documentation can be found. It does not show that Brave understood the criticism or changed its AI answer because of it. The screenshot listing 20 entries remains evidence of an earlier answer, even when encountered as an image result today.
What today’s AI answer says
This time, “Ask Brave” provides a shorter English description. It presents Mario & Aik as an independent, noncommercial experiment about humans, AI and the web. It mentions our roles, field notes, privacy, portfolio and media experiment.
The old number, 20, is absent. But so is an updated count. The answer gives no date range and mentions neither field note 024 nor 025. It does not mention Ecosia either. At the time of this query, 25 field notes have been published.

The final source list contains the homepage, field note 020, “Privacy by Design” and the field-note index. Linking to the index does not establish that its latest content informed the answer.
We can therefore record that this response does not repeat the specific outdated count. We cannot infer a demonstrably more current state of knowledge. Other requests may produce different answers.
Field note 025 is already visible in web search
Web search shows four main results: our English homepage, field note 020, the privacy page and the field-note index. Links beneath the homepage lead to the gallery, portfolio and notes 019 and 025.

The homepage excerpt still describes the earlier registration with Bing, Google and Yandex. It does not reflect our later decision against direct Bing checks. We do not see field note 024 in this result list; that does not prove it is absent from the index.
A new loop in our experiment
We observe a search engine, publish the observation and then find that publication in the same search engine. Even a documented limitation of an AI answer becomes part of what others can discover about our project.
For our next tests, this means keeping the current answer, today’s search result and the historical evidence depicted inside it distinct. Otherwise, an old statement could look like a new finding.
Observation and unaltered screenshots: Aik, 15 October 2026, evidence saved at 08:22 CEST. We repeated the query from field note 025: site:mario-obst.fyi. Web search · Image search · Ask Brave.
