Case Study — AI Search / Entity Visibility (our own names)

We Asked ChatGPT, Gemini and Perplexity About Ourselves. Each Knew Something Different

By Sai Teja (tejalogs) · 3 November 2026

The short answerOn 3 November 2026 we asked ChatGPT (with web search on), Gemini and Perplexity the same four questions about our own names. Perplexity correctly explained the handle tejalogs. Gemini correctly described The Social Algorithm. ChatGPT described the agency correctly but could not find the handle. Only Perplexity connected the founder to the agency, and only when asked for the bare handle. For the bare name "Sai Teja", all three returned other people, which we expected.

How we ran it

Four queries, three assistants, one run each: tejalogs; Sai Teja; Sai Teja AI marketing Vijayawada; and The Social Algorithm agency. ChatGPT ran in a temporary chat with personalisation switched off and web search on. Gemini ran in a temporary chat. Perplexity ran signed out. It is a single snapshot, and these systems vary from run to run.

What each assistant answered

tejalogs. Perplexity: correct, citing GitHub, LinkedIn and our site. ChatGPT: no match, returning a logistics company, a payments app and an old forum username. Gemini: no match, offering unsourced guesses.
Sai Teja. All three: other people with the same name, or a request to clarify.
Sai Teja AI marketing Vijayawada. Perplexity and ChatGPT found the founder through LinkedIn but did not mention the agency. Gemini named three other people.
The Social Algorithm agency. Gemini: correct, from our site. ChatGPT: correct, though still quoting three case studies we had removed that day. Perplexity: no match, offering a research group at the MIT Media Lab and two other agencies.

What we think is going on

These are hypotheses, not findings. ChatGPT's search appears to miss the bare handle, and Bing's own data shows our About page had last been crawled in August, so a stale index is one candidate. Perplexity seems to lean on GitHub and LinkedIn. The phrase "social algorithm" collides with an unrelated research group, and the name "Sai Teja" is shared by many people.

What was wrong in our own house

Our site called the founder "Teja" in bylines and structured data, while the handle, LinkedIn and the About page said "Sai Teja". In an earlier test the same day, ChatGPT had attached another person's full name and photo to the handle. We cannot prove the inconsistency caused that, but it is the kind of gap an assistant fills with a guess.

What we changed on 3 November

One published name across the site, Sai Teja, with the handle tejalogs. One identity sentence, repeated word for word on three new pages: who is tejalogs, what The Social Algorithm is and is not, and whether it is legit. Our LinkedIn company page is linked into the site's structured data, and stale pages were resubmitted to Bing.

What this does not prove

Twelve answers from one afternoon is a small sample. Assistants change their answers between runs, and we cannot show that any one change will move any one answer. We will not know whether the changes helped until we run the same tests again.

What we will do next

On 1 December 2026 we will re-run the same twelve tests and publish the comparison, including if the results are worse.

FAQ

Which AI assistant knew about tejalogs?

On 3 November 2026 Perplexity described the handle correctly, using GitHub, LinkedIn and our site. ChatGPT's search and Gemini did not find it.

Why can't an AI assistant find Sai Teja?

Sai Teja is a very common name, so a search for the bare name returns other people on every assistant we tried. The handle tejalogs or a descriptive query such as AI marketing in Vijayawada works better.

Will you repeat the test?

Yes. We will re-run the same twelve tests on 1 December 2026 and publish the comparison, including if the results are worse.