Asking an AI who you are is an odd experience. It can recognise your work in one answer and give you somebody else's career in the next.
On 20 October 2026, I used GEDI to run four questions across 12 AI engines. The questions covered identity, expertise, possible collaboration, and gaps or contradictions in the available information.
What the first run showed
The recorded responses included correct associations between Nicolas Chan, Neekchan, and communications work at The Hoffman Agency. They also included confusion with unrelated software-engineering and digital-marketing profiles.
In one sequence, ChatGPT connected the identity correctly in its earlier responses, then described a different front-end and accessibility practitioner in the collaboration question. Claude and Perplexity also returned unrelated profiles in parts of the run. These are observations from that batch, not permanent assessments of those products.
That inconsistency matters. A correct answer to “who is this?” doesn't mean a later answer about their work will stay attached to the same person.
How the check was run
| Setting | This baseline |
|---|---|
| Run date | 20 October 2026 |
| Questions | Four, with one pass per engine |
| Engines | ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek, Qwen, Volcengine, Kimi, Ernie, GLM, Hunyuan |
| Stored result rows | 48 |
| Returned text | 47; not all were substantive answers |
| Failed result | One Hunyuan response returned no usable text |
The batch used API configurations in GEDI. Search and grounding varied between engines; these results shouldn't be treated as a reproduction of every consumer app. At least one returned response announced research without supplying findings. Raw exports are not published with this note.
This was one small baseline. It doesn't establish an accuracy ranking, a visibility score, or what everyone searching for me will see.
What I'm changing
I'm putting a clear About page on this domain, connecting my official profiles, and keeping references alongside the things I publish. The identity note spells out the names and aliases.
The practical question for a later run is whether the same questions retrieve the right person more consistently, and whether they cite useful evidence. That comparison hasn't happened yet.
The boring part that matters
Useful information needs to be readable and attributable. Google says its existing search fundamentals also apply to AI features. Its ProfilePage guidance describes structured data for a page about a person. OpenAI separately documents its search and training crawlers.
Those are implementation references. They don't promise that an engine will index a page, cite it, or get the answer right. I'll keep the outcome separate from the intention.