How Agencies Are Adapting to ChatGPT Ads (And Why Most Aren't Ready)

The natural move for most agencies when ChatGPT advertising launched was to repurpose their existing practices. Import the brief architecture from search. Adapt the creative templates from Meta. Scale what worked somewhere else.

This approach has consistently missed the format entirely.

The frameworks that don't transfer

Search advertising is built around keywords. The keyword is the unit of organization. Everything — ad groups, match types, negative keywords, Quality Score — serves the keyword. An account structure that works for search is fundamentally organized around the wrong thing.

ChatGPT advertising has no keywords. It has no account hierarchies. It has no match types or Quality Scores. The organizing unit is the conversation — and the conversation is something that exists inside OpenAI's infrastructure, not something the advertiser controls.

Where traditional practices break down

The creative templates that work in search are built for scanning. A headline designed to land visually on the SERP requires a certain kind of brevity and specificity.

ChatGPT ads are read, not scanned. They appear after a user has just spent mental effort on a topic. The argument has to hold up to reading.

What the winning agencies are doing differently

The agencies winning on ChatGPT are rethinking the brief architecture itself. Instead of writing a single ad asset and distributing it across contexts, they're mapping ad copy to specific conversational states.

The brief includes: the specific conversational context the ad is likely to appear in; the mental state of the user at that moment; the specific problem the user was reasoning about; what new piece of information would be useful right now.

The measurement infrastructure challenge

Most agencies still think of measurement as clicks and conversion rates. For ChatGPT, that's incomplete.

The better framework: impression quality (is the ad appearing in genuinely relevant conversations?); click quality (are users clicking from a state of having already done research?); and post-click behavior (are users spending meaningful time, moving through steps, signaling purchase intent?).

The briefs that work

The best ChatGPT briefs do five things:

First, they identify the specific conversational triggers. Not "anyone researching software" but "someone 6 months into evaluation, comparing three known options, building a decision matrix."

Second, they write the ad as a specific piece of new information the conversation probably needs. Not a pitch. Not a claim. A fact that answers an unstated question.

Third, they design the landing page as the next line in the same conversation. Not a homepage. Not a generic funnel. Continuity.

Fourth, they test creative variations against specific conversational states, not just broadly.

Fifth, they measure things that matter: not CTR or impressions, but impression quality and downstream behavior.