SponsorSight

A browser extension that flags AI chat answers carrying a paid placement the platform never labeled.

Problem

ChatGPT now puts a sponsored answer inside roughly a quarter of its Free and Go-tier responses, and Google AI Mode shows ads on close to 30% of ad-eligible queries. OpenAI says paid placements "stay clearly separate" from the answer itself, but nobody outside OpenAI has checked whether that separation holds up in a live conversation, and Google doesn't give even its own advertisers placement-level reporting for AI Mode or AI Overview. Ask ChatGPT which credit card or mattress to buy, and you're trusting a label you have no way to verify was applied.

Target user

People who've made "just ask the AI" their default shopping research step. They used to read five Reddit threads before buying anything over $50; now they ask one question and act on the top answer. Job-to-be-done: tell me when the recommendation I'm about to act on had money behind it, even if the platform didn't flag it.

MVP scope

• Content script for chatgpt.com, Google AI Mode results, and perplexity.ai that reads each rendered answer block and records any native "Sponsored" or "Ad" marker the platform shows.
• A webRequest listener that timestamps outbound calls to known ad-serving and ad-tracking domains fired while an answer is generating or rendering.
• A matching pass that lines up ad-network timestamps against rendered answer blocks and flags any block with ad-network activity but no visible label, next to the paragraph itself.
• A toolbar badge showing how many answers got flagged in the current session.
• A local-only export (JSON or CSV) of flagged answers with timestamp and matched domain, so a user has a record before filing anything with the FTC or a state AG.
• A settings panel to toggle each platform independently, since the ad domains and DOM structure differ per site and each needs separate upkeep.

Monetization

Freemium. Free tier covers ChatGPT only, with the badge counter and session-only history. $5/month unlocks Google AI Mode and Perplexity, persistent history with export, and a weekly digest email of flagged answers.

Why now

Similarweb's August 17, 2026 data showed 26% of ChatGPT responses on the Free and Go tiers already carry a sponsored ad, and close to 30% of ad-eligible Google AI Mode queries do too. OpenAI's head of global ads solutions told press the format keeps paid placements "clearly separate" from the answer, but the same week's coverage pointed out that Google AI Mode and AI Overview don't give even advertisers a way to check where their own ads ran, a gap that applies to ordinary users too. The labeling promise exists. The ability to check it from outside the platform doesn't.

Risks & open questions

• The ad-serving domains behind ChatGPT's ad pilot aren't documented anywhere since the program only started this year, so the domain list has to be reverse-engineered by watching traffic, and it will drift as OpenAI iterates.
• DOM structure on all three target sites changes often and without notice; the extension can break silently until someone notices flagged counts dropped to zero.
• Will anyone pay $5/month to learn an AI answer was sponsored, or does the free ChatGPT-only tier cover the entire addressable audience? People annoyed enough to care might just switch assistants instead of installing a watchdog.
• Correlating "an ad-network request fired" with "this paragraph was paid for" is a heuristic, not proof. A concurrent ad-network call could be an unrelated tracking pixel, and every flag needs to say so instead of implying certainty.
• OpenAI or Google could treat request logging and DOM scraping of their own products as a terms-of-service violation and try to detect or block the extension.

Next step

Spend a week manually watching network traffic on chatgpt.com and Google AI Mode while asking product-recommendation questions, to confirm ad-network calls actually correlate with sponsored answer blocks before writing the matching logic.

Sources

• https://ppc.land/similarweb-opens-ai-ad-data-as-26-of-chatgpt-replies-carry-sponsored-ads/ — Similarweb's August 17, 2026 data: 26% of ChatGPT Free/Go responses carry a sponsored ad, ~30% of ad-eligible Google AI Mode queries show ads, and neither platform gives advertisers placement-level reporting.
• https://abc17news.com/stacker-ai/2026/09/11/ads-have-arrived-in-chatgpt-what-the-future-of-ai-search-could-mean-for-brands/ — September 11, 2026 piece on OpenAI's ad rollout, including its "clearly separate" claim and an analyst's warning that clumsy ad placement could push users to competitors.