Why People Respond to Ads Differently Inside a Conversation
By The InPromptAds Team 12 min read
Most digital advertising is in the business of interrupting people. You're watching a video, reading an article, scrolling a feed — and an ad lands in the middle of it, asking for attention you weren't offering. The entire architecture of digital advertising has been built around this adversarial dynamic, and the entire history of ad avoidance has been the consumer's response to it.
ChatGPT advertising is structurally different — not as a matter of degree but in kind. And the reason why has nothing to do with the ad itself. It has everything to do with what happens to a person psychologically in the seconds before the ad appears.
The reason you hate most ads
There's a concept in consumer psychology called psychological reactance. It's the motivational state that kicks in when you perceive your freedom or attention being taken without consent. In advertising contexts, it manifests as irritation, avoidance, and in its most developed form, a trained reflex to mentally route around anything that looks commercial.
Research published in the Journal of Advertising identified the two variables that determine how intrusive an ad feels: how relevant it is to what you're currently thinking about, and how deeply engaged your mind is when it lands. The more cognitively occupied you are with something unrelated, the more invasive any commercial interruption feels. That's why pre-roll ads are so consistently resented — you came for the video, your mind is primed for the video, and someone is demanding thirty seconds of your attention before you can have the video. Reactance is basically guaranteed.
Banner blindness — documented since the late nineties and consistently replicated since — is the long-term result of this dynamic at scale. After years of ads arriving in predictable positions with no relevance to the task at hand, the brain simply learns not to look there. It's not aesthetic fatigue. It's a trained avoidance response to chronic mismatched interruption.
Search advertising reduced the mismatch significantly. Linking an ad to a declared keyword is as close as digital advertising has come to serving something relevant at the right moment. But it didn't eliminate the interruption. The user came for results, not pitches. The ad is still an unsolicited presence in a context the user didn't request.
Here's what makes ChatGPT structurally different. The ad doesn't appear while you're doing something else. It appears after you've just received an answer to a specific question — precisely the question the ad was contextually matched against. The user isn't being interrupted. They're being presented with something relevant at the exact moment when their mind has just finished processing one thing and hasn't started the next. That pause — cognitively light, topic-focused, attention still present — is the window every other ad format has been trying to reach for thirty years.
Your brain at the moment the ad appears
Cognitive load across one ChatGPT interaction
1. Types the question
Load rising
2. Reads the response
Load at peak
3. Natural pause
Load releases
4. Sponsored card appears
The reasoning window
There's a specific cognitive state that makes ChatGPT ads different from both search and social, and it's worth being precise about it.
When someone is in the middle of reading a ChatGPT response, their working memory is at high capacity. They're parsing a multi-sentence answer, holding context from earlier in the conversation, evaluating whether the response addresses what they asked. Cognitive load is near-peak. That's actually a poor moment to introduce a new stimulus — which is why the chat_card appears after the response, not inside it. By the time the ad is visible, the cognitive load has released. The answer has landed. The mind is momentarily clear.
This is not an arbitrary design choice. It's the ideal neurological window for a new input to arrive without triggering the overload response that makes people dismiss things without reading them.
And because the user was just actively reasoning about the topic the ad addresses — not passively scrolling past it, not distracted by a feed of unrelated content — they arrive at the ad in what persuasion researchers call a high-elaboration state. In plain terms: their critical faculties are engaged, they care about the topic, and they have the mental bandwidth to actually evaluate what the ad is saying. This is the opposite of the low-attention peripheral processing that display advertising relies on.
The practical implication of this is significant and cuts both ways. An ad that makes a genuine, specific, well-reasoned case in this context will be processed more carefully and stick more durably than the same argument on any other format. An ad that's vague, generic, or off-register won't just be ignored — it will be evaluated and found wanting by a mind that's actually paying attention. There's nowhere to hide behind production value or peripheral cues when the user is actively reading.
The targeting advantage nobody talks about
Every form of digital advertising prior to this one has been working from incomplete information. Behavioral targeting observes what you've browsed, clicked, and purchased, and makes a prediction about what you currently want. It is, at its best, educated inference from historical data. It tells you what someone was interested in and guesses at what they might want now.
Conversational advertising is something different. It's not inference. It's observation.
The user just described their specific problem, in their own words, in the present tense, to the system that matched the ad. "What are the main differences between Rippling and Gusto for a 150-person company with contractors in three countries?" is not a signal that can be constructed from browsing behavior. It's a direct declaration of need, context, scale, and decision stage — more targeting information in a single query than behavioral systems can typically accumulate over weeks of passive observation.
This connects to a finding in conversational AI research that doesn't get nearly enough attention in advertising discussions: people disclose significantly more to AI interfaces than to any equivalent static input. Studies on chatbot interaction consistently show users volunteering personal, specific, and sensitive information in conversational contexts that they wouldn't enter into a form or a search box. The social logic of dialogue — that you give context, the other party responds to it, and you both converge on the same understanding — activates a norm of reciprocal disclosure. Users follow that norm automatically, often providing far more detail about their actual situation than the ad system would have any other way of knowing.
The gap between behavioral prediction and conversational observation is not a marginal one.
The difference between a salesperson who researched your LinkedIn before the call and one who just spent twenty minutes actually listening to you describe your problem.
That's the targeting advantage conversational advertising has over everything that came before it.
Why trust transfers to the ad
Here's the mechanism that's underappreciated even among people who think seriously about this format.
By the time the sponsored card appears, the user has already formed a relationship with the interface. Not a deep relationship — but a transactional one in which the AI has demonstrated that it understands them, responds to their specific question, and operates in their interest. That experience creates what psychologists call a halo effect: positive affect generated by one stimulus transferring to adjacent stimuli.
The ad doesn't arrive as a stranger. It arrives in the context of an interaction where the interface has already earned a degree of credence. The user isn't consciously thinking "I trust ChatGPT, therefore I trust this ad." But the cognitive context in which the ad appears is one of established cooperation — and that context shapes how the commercial signal is received, whether the user is aware of it or not.
Research on conversational AI and parasocial trust — published at ACM FAccT in 2024 — documents how chatbots generate trust through linguistic conventions that mirror human social behavior: acknowledgment, turn-taking, personalised responses, affirmations. Users apply social trust norms to these interfaces even while intellectually knowing they're not human. The practical result is that a recommendation appearing at the end of a ChatGPT interaction carries a different weight than the same recommendation appearing cold on a display network.
Call this the dialogue dividend: the persuasive advantage that accrues to a commercial message because it arrives at the conclusion of a cooperative exchange, rather than from an anonymous network into an unprepared mind. A Princeton study in 2024 demonstrated the scale of this effect — users of conversational AI selected sponsored products at nearly three times the rate of search engine users when products were promoted conversationally. That study modelled covert influence, which OpenAI explicitly prohibits. But it quantifies the underlying mechanism. The dialogue dividend is real, and it operates even in a fully transparent format.
What you actually have to do differently when you write the ad
Everything above leads to one creative conclusion that most agencies are currently ignoring because they're repurposing Google and Meta briefs for ChatGPT.
An ad appearing in a high-elaboration, post-conversation, trust-primed context will be read, not scanned. Which means the argument has to hold up to reading.
Here's what that looks like in practice. Two versions of the same ad — same product, same placement, different understanding of the format:
Version A (search brain): "Streamline your HR workflows. Trusted by 10,000+ companies. Get started free."
Version B (conversation brain): "Built for distributed teams with contractor compliance across jurisdictions. Free for up to 50 users. No implementation fee."
Version A is a perfectly competent search ad. It would survive keyword matching and pass creative review. In a ChatGPT conversation where someone just asked a detailed question about managing contractor payroll across multiple countries, it reads as generic noise — because the user's mind is engaged with a specific problem and the ad isn't speaking to it.
Version B addresses the conversational state. It reflects back the specifics of the user's situation, provides a concrete differentiator that's relevant to what they were researching, and removes a friction point they probably have (cost of switching). It treats the user as someone who just told you exactly what they need — because they did.
This is the creative discipline that conversational advertising demands. Not cleverness. Not disruption. Specific relevance, delivered in the register of a competent referral. B2B-focused practices like InPromptAds build this into the brief architecture itself — mapping ad copy to the specific conversational state it's likely to appear in, rather than writing a single asset and distributing it across all contexts. The format rewards that level of specificity because the user is capable of evaluating it. And it penalises vagueness in a way that lower-attention formats simply don't.
The limit you can't push past
The dialogue dividend has a ceiling, and understanding it is as important as understanding the dividend itself.
The trust that makes this format work is contingent on one thing: the conversation being real. The user extends goodwill to the interface because the interface has been genuinely helpful. That goodwill is not infinite, and it is not unconditional. Research on parasocial AI relationships consistently finds that when users perceive the trust being exploited — when the cooperative dynamic starts to feel like a setup — the backlash is disproportionate to the offence. The same emotional investment that creates the halo effect can reverse it sharply.
This is why OpenAI's Answer Independence principle — the commitment that ads will never shape the AI's response — is not just regulatory compliance. It's the structural protection of the very thing that makes the format valuable. A ChatGPT ad that feels genuinely congruent with a helpful answer builds on the dialogue dividend. An ad that makes the user feel the whole conversation was a commercial setup destroys it.
The format works because it doesn't try to be more than it is: a contextually relevant commercial signal arriving at the right moment in an exchange that already happened on its own terms. The moment advertisers start treating the conversation as a delivery mechanism for the ad rather than an independent interaction that earns the right to one, the psychology inverts entirely.
The dialogue dividend is not a trick. It's a consequence of doing the format honestly.
Quick answers
Why do ads inside ChatGPT feel less intrusive than banner or pre-roll ads? Because of when they arrive. Research in the Journal of Advertising identifies two variables that determine how intrusive an ad feels: how relevant it is to what you are currently thinking about, and how cognitively occupied you are when it lands. A pre-roll ad interrupts a mind primed for something else. A ChatGPT ad appears after the response has landed, when cognitive load has released and the topic is still active.
What is the reasoning window in ChatGPT advertising? The moment after a ChatGPT response ends and before the user forms a conclusion. Working memory is at high capacity while a user reads an answer, which is a poor moment to introduce a new stimulus. The sponsored card appears after the response rather than inside it, when load has dropped and attention is present.
What is the dialogue dividend? The persuasive advantage a commercial message gains from arriving at the end of a cooperative exchange rather than from an anonymous network into an unprepared mind. A 2024 Princeton study found users of conversational AI selected sponsored products at nearly three times the rate of search engine users when products were promoted conversationally. That study modelled covert influence, which OpenAI prohibits, so it quantifies the mechanism rather than the permitted format.
Does psychological reactance apply to ChatGPT ads? Less than to interruptive formats, and not zero. Reactance is the state that kicks in when attention is taken without consent. Placement after the answer, topical relevance and a visible sponsored label all reduce it. Ads that feel incongruent with the answer still trigger it.
Why does OpenAI's Answer Independence principle matter commercially rather than just legally? Because the trust the format borrows is the thing being sold. If users came to believe a conversation had been shaped commercially, the credibility that makes the placement valuable would go with it. Answer Independence protects the mechanism, not just the policy position.
Does this mean ChatGPT ads perform better than search ads? It does not. This is an argument about how the format is received, not a performance claim. OpenAI states there are no performance benchmarks across advertisers, industries or campaign types, so no comparative claim in either direction is currently supportable.
Related reading
- What Are ChatGPT Ads? — the format and targeting this piece assumes.
- ChatGPT Ads vs Google Ads vs Meta Ads — why creative that works on Meta underperforms here.
- The 10 Best ChatGPT Ads Agencies for B2B Brands — who is building creative for this format.
- ChatGPT Ads Troubleshooting — when the creative is fine and the delivery is not.
- Prompt-aligned ad copy and landing continuity — how we operationalise all of this.
- Glossary and ChatGPT Ads.
Primary sources: Ads in ChatGPT: The Basics · Ad policies
InPromptAds is a B2B specialist agency focused exclusively on ChatGPT native ad placements, operating as part of the Newtation Group.
Field kit
Tools
Site
OpenAI (primary)
The InPromptAds Team
Writes about AI-native advertising and the shift from search to conversation.
