AI Agent Advertising Faces Attribution Crisis as Market Takes Shape
Startups race to measure ads served to chatbots and AI agents, but no standard has emerged for tracking whether these new formats actually work.
As publishers begin serving advertisements directly to AI agents rather than human readers, the industry confronts a fundamental measurement problem: how do you prove an ad worked when no person ever saw it?
The challenge has spawned a wave of startups pitching attribution solutions, but the market remains far from consensus on standards. Most approaches are still proof-of-concept, and measurement remains largely opaque.
How AI agent ads work
The emerging ad formats target AI agents as they scrape publisher websites to answer user queries. When ChatGPT or similar tools crawl a webpage, specialized ad platforms can inject sponsored content into the HTML or markdown that the agent retrieves.
OpenAds, for instance, generates unique referral codes for products mentioned in AI responses. If a user asks ChatGPT about running shoes, the platform inserts an ad with a discount code when the agent scrapes relevant content. The company tracks conversions when users redeem those codes, according to co-founder Steven Liss.
Oasy takes a different approach, working with native ad platforms to inject text-based sponsored ads when AI crawlers are detected through content delivery network integrations. Co-founder Choy Travers said the goal is to "influence the chance" that brands appear in AI outputs, though he declined to name current publisher or advertiser partners.
Time's recent agent advertising initiative uses FAQ-style sponsored ads in markdown pages, with measurement partner Mobian tracking AI bot traffic and testing how often brands surface in chatbot responses.
The attribution gap
Traditional digital advertising follows a clear path: impression, click, conversion. AI agent advertising breaks that model. An agent must retrieve the ad, incorporate it into a response, and influence what a user ultimately purchases—all without the consumer seeing or clicking the original advertisement.
"How do you attribute that the ad was actually seen by an agent and then influenced the response?" said Brendan Norman, founder of contextual advertising firm Classify. "It's not as straightforward as a clickthrough or a checkout."
Publishers can track when AI agents crawl pages containing ads, but that doesn't confirm the agent processed the sponsored information or surfaced it to users. Attribution becomes even murkier because agents typically synthesize information from dozens of sources when answering a single query.
"That's where everyone is stuck at the moment," Liss said.
Some firms propose using analytics companies to sample synthetic prompts and measure brand visibility before and after ad campaigns. But Liss acknowledged that approach won't scale the market. Advertisers want bottom-of-funnel attribution tied to actual revenue before committing meaningful budgets beyond experimental spending.
Why it matters
The measurement impasse threatens to stall a potentially significant new revenue stream for publishers as traditional search traffic declines. Without reliable attribution, advertisers have little reason to shift budgets from proven channels to experimental AI agent formats. The IAB is drafting attribution frameworks for agentic advertising, but industry-wide standards remain months or years away. Publishers and ad tech vendors who solve attribution first could capture outsized market share as AI agents become primary information sources.
These details were first reported by Digiday in its Media Briefing.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call