Graphite & Similarweb (2026) — State of ChatGPT Ads

Summary:

This report analyzes how ads appear in ChatGPT, who’s buying them, what creatives they run, and what’s working so far.

Published by Graphite (a growth agency) with Similarweb (digital intelligence data), it aims to map the channel early and turn it into an advertiser playbook — both with clear commercial incentives.

Data sources:
• Similarweb: panel-based reach, impression share, CTR, and ad position across nine live ChatGPT ad markets (as of Sept 2026). Shares percentages only, not panel size.
• Graphite: 95K+ US ads from 500+ advertisers (Apr–mid-Aug 2026), plus 70K+ creatives, 50K+ headlines, and 8K+ landing pages crawled and categorized. It also classified ~9K of the top 10K US advertisers.
• Small cross-channel check: Meta Ad Library + Google Ads Transparency Center for 10 advertisers active on all three.

Key findings:
• In the US, ads reached 43% of ad-eligible users; global CTR more than doubled since launch. Highest CTRs: Brazil, Mexico, Japan.
• Earlier placements win: first four turns had 48% higher CTR, and 51% of impressions landed in the first three turns.
• Advertisers look in test mode, rotating in and out rather than spending steadily.
• B2B SaaS leads by impression share; retail/ecommerce lag.
• Biggest advertisers push “act now” offers and use more creative variants and landing pages.
• 98% of headlines were net new (not copied from other channels). Brand mentions in prompts can trigger competitor ads.
• Most advertisers use ≤5 landing pages; larger ones use more. Feature/solution/pricing pages dominate.
• In the 10-advertiser sample, about half ran more ads on ChatGPT than Meta, but Google Search still had far more overall volume.

Caveats:
Vendor-led study with mixed methods (panel + Graphite crawl + tiny cross-channel sample). Similarweb doesn’t disclose panel size. Some results are explicitly early/directional. Landing-page classifier is imperfect (84.7% accuracy).