PUBLIC EXPERIMENT

AI Search Lab

A public experiment to find out whether AI tools such as ChatGPT, Perplexity, Gemini, Google AI Overviews, Claude and Copilot mention me, link to this website, and describe me correctly.

6 AI tools

30 questions

3 runs each

Checked weekly

AI Search Network Signal GraphA central node representing sauravsahoo.com linked by connecting signal lines to six AI tools.CopilotChatGPTPerplexityGeminiGoogle AI OverviewsClaudesauravsahoo.com

Purpose

I want to find out, with evidence, how AI tools describe me and this website, and whether the work I do on the site changes that. I will publish the method and the results, good or bad.

Methodology

1. Ask

I ask the same list of questions every time. The list is fixed once the first round starts.

2. Record

For each answer I note three things: was I mentioned, was my website linked, and was the description correct.

3. Repeat

Each question is asked 3 times on each AI tool, every week, because AI answers change from one run to the next.

4. Compare

Results are shown as percentages. The first round is the baseline, so later changes can be compared with it.

A change after an update does not prove the update caused it.

Scope

What is tracked

Six AI tools:

  • ChatGPT
  • Perplexity
  • Gemini
  • Google AI Overviews
  • Claude
  • Copilot

Five question groups:

  • Brand8 questions
  • Local searches7 questions
  • India9 questions
  • Role and generic4 questions
  • AI search topics2 questions

What this does not cover

  • Paid ads
  • Personalised results for individual accounts
  • Guarantees about future visibility

Results can differ by location and account, and AI tools change often.

Current Stage

Stage 1 Done

Design

The method and question groups are published.

Stage 2 In progress

Baseline run

The first round of questions, before any further changes.

Stage 3 Upcoming

Weekly tracking

The same questions, every week.

Stage 4 Upcoming

First review

A written summary of what the data shows.

Data

Mentioned

–

Linked

–

Described correctly

–

No results published yet. The first numbers will appear here after the baseline run.

Experiment Chronology

2026-10-31Done

Experiment design and question groups published

Planned

Baseline run

Planned

Weekly tracking begins

Planned

First review published

EXPLORE MORE

Want to learn more or discuss AI search?

Last updated 1 November 2026