Public research preview

Explore 36,000 ads observed in ChatGPT

9,661 advertisers, 22,713 associated prompts and 383.5K captured appearances in the current snapshot.

Independent research libraryNot operated by OpenAI. Counts describe captured source appearances—not impressions, spend, reach or campaign performance.
Observed ads
36,000
Advertisers
9,661
Buyer prompts
22,713
Captured appearances
383.5K

Counts refresh from the active catalog rather than a fixed marketing total. As new ads, advertisers and supported industries are added, the public snapshot and discovery files update from the same source dates.

Updated

Industry directory

Browse markets with sufficient data.

New industries appear automatically after they meet the public quality threshold. Other and low-volume categories remain outside the index.

Public creative sample

A useful preview, with a deliberate limit.

Only short, dated examples are exposed here. Full ad copy, associated prompts, source evidence, advanced filters and export stay behind Pro access.

R
Robinhoodrobinhood.com
Observed
Sponsored

Let your AI agent trade

Market access for AI agents w/ Robinhood

Topic
Cloud & DevOps
Source last seen
2026-08-23
Captured appearances
4,948
I
Igeniei-genie.ai
Observed
Sponsored

Augmented intelligence for consumer insights

i-Genie is provides next gen Consumer Insights without surveys or focus groups.

Topic
Cloud & DevOps
Source last seen
2026-08-13
Captured appearances
2,973
Z
ZoomInfo Technologies Incgtm.ai
Observed
Sponsored

One API. Every GTM Signal.

One endpoint for AI agents and GTM workflows. Pay for what you use.

Topic
Cloud & DevOps
Source last seen
2026-08-23
Captured appearances
2,540

ChatAdsWatch Pro

Search the full snapshot around your own market.

Unlock complete creatives, buyer prompts, competitor views, saved reports and credit-backed keyword or classification analysis.

Open the app preview
01

How is this library collected?

The catalog preserves captured ad records, associated prompts and upstream observation counts. Inferred industry, topic and intent labels are stored separately from source evidence.

02

How complete is it?

It is a dated sample rather than a census of everything shown in ChatGPT. Missing geography and uncertain classification remain explicitly unavailable.

Read definitions, limitations and classification rules →