ContextHints Intel · Free to start
Ad intelligence for ChatGPT Ads, with the evidence attached.
See which ads run in ChatGPT, which company is behind each one and what it sells. Every inferred context hint shows how many ads support it, from how many independent sources, and how old the data is.
Free account, no credit card required. ChatGPT coverage today. Other AI engines are added one at a time, each after its own compliance review.
Inferred context hint
ModerateTeams comparing tools that track brand mentions in AI answers
- Advertiser
- Example Analytics Co.
- Niche
- Marketing › AI search optimization
- Ads
- 7
- Sources
- 3
- Days seen
- 12
- Effective n
- 5.8
Why it's different
Built to be checked, not taken on trust.
Ad data for AI engines is new, sparse and easy to over-read. These five rules shape everything you see in the product.
Entity resolution you can check
Each ad is resolved to the company behind it through its registrable domain, and every entity lists the aliases and domains behind it. Redirect and tracking domains sit on a separate list and never become advertisers. Uncertain merges go to a human reviewer.
Product-based niches
Advertisers are classified by what they sell, read from the ad and its landing page. The topic of the conversation an ad appeared in is kept as a separate field, so a placement never decides the category.
Evidence-weighted confidence
Confidence is computed in code from the evidence: ads, independent sources, days observed and effective sample size, pulled toward caution when evidence is thin and decayed as it ages. The band thresholds are early-access placeholders until they are fitted on labelled data, and the app says so.
Honest freshness
Each capture source shows the date its data runs through and its measured lag. A source that falls behind is flagged as stale, not hidden.
Compliant capture
No automation of consumer AI accounts, ever. Data comes from analyst capture, licensed or donated imports, an authenticated ingestion API, and your own ad accounts.
How it works
From captured ad to usable intelligence.
Step 1: Capture
Ads arrive through four channels: an analyst capture form, CSV or JSON imports of licensed or donated ads, an authenticated ingestion API, and data from ad accounts you manage. Tracking parameters are stripped against an allowlist on the way in. Ad click URLs are never followed.
Step 2: Enrichment with Jev
TypeSafe's Jev model answers narrow, typed questions about each new creative: is this a real sponsored ad, which advertiser is it, which product niche fits. Answers are yes/no, pick-one or graded judgments with probabilities. Jev never writes text. Code does the counting, dates, domains and evidence math, and low-confidence cases go to a review queue.
Step 3: Intelligence
Browse the ad library, advertiser profiles and the niche tree. Inferred context hints are assembled from verbatim ad text and carry their full evidence, so you can see why each one scored the way it did.
Compliant capture
Compliance is a design constraint, not a disclaimer.
We'd rather have less data than data we can't stand behind. These rules are part of how the product is built, not a disclaimer added afterwards.
How data gets in
Analyst capture
Manual capture by our analysts on their own accounts during normal use: no scripted prompt lists, no ad clicks. Each capture records the analyst's acknowledgement of the capture procedure and counts against a daily cap.
Imports
CSV or JSON files of licensed or donated ads, each with a provenance reference and a rights attestation. Every record is tagged with where it came from, and a customer's own imports stay private to their workspace.
Ingestion API
Authenticated with per-workspace keys. A consented browser-extension panel is designed but stays switched off until counsel and OpenAI have signed off.
Your ad accounts
Managed-account data stays inside your workspace. It is never pooled or shown to anyone else, in line with OpenAI's Advertising Terms.
What we never do
- Automate ChatGPT or any other consumer AI account
- Send prompts to AI engines to harvest ads
- Rotate proxies or spoof browser fingerprints
- Solve CAPTCHAs or evade bot detection
- Click ads or follow ad click URLs
- Store the text of the conversation an ad appeared in
To classify what an advertiser sells, we may read their public landing page: HTTPS only, robots.txt respected, no cookies or logins, and rate-limited per site. At most, an analyst picks a topic category and adds a short note in their own words, never the conversation's text.
Start free. See the evidence for yourself.
Create a free account, verify your email and you get your own workspace. No credit card required. Invite your team when you're ready.