ChatGPT Ads: Lost, Confusing & 🤯 💸

ADVERTISING IN CHATGPT: A DETAILED ANALYSIS
Searchable’s recent analysis of over 11,000 ads served within ChatGPT conversations between July 4th and August 4th, 2026, reveals significant challenges for advertisers utilizing this emerging platform. The core of the investigation involved meticulously pairing each ad with the conversation in which it appeared, subsequently grading the ad’s relevance based on three distinct bands. The findings highlight a substantial disconnect between advertiser intent and user engagement, revealing a concerning level of misalignment within the ChatGPT advertising ecosystem. Specifically, a striking 33% of ads were deemed entirely unrelated to the conversation, demonstrating a lack of connection to the user’s immediate query. Only 27% of ads achieved a direct match, precisely reflecting the product or service the user was actively asking about. The remaining 40% fell into the category of contextual matches, appearing in threads where the topic had been previously raised but not directly tied to the current question. This broad categorization underscores the difficulty advertisers face in accurately targeting users within this dynamic conversational environment.

RELEVANCE BANDING AND INDUSTRY VARIATIONS
The methodology employed by Searchable—grading ads across three relevancy bands—provides a valuable framework for understanding the performance of ChatGPT advertising. However, the analysis reveals significant variations in relevancy rates across different sectors. The overall mismatch rate, when viewed through the lens of conversation rather than individual ad counts, indicates a substantial issue: 40% of ad-carrying chats contained at least one irrelevant ad, and in 28% of instances, every single ad served was disconnected from the user’s topic. This highlights a critical operational problem for advertisers. Furthermore, the analysis demonstrated a strong correlation between sector and relevancy. Data brokers and background checks yielded the most relevant targeting at 50%, followed by travel (43%) and automotive (42%). Conversely, marketing and B2B services saw a significantly higher mismatch rate at 47%, and software and SaaS ads also struggled with relevancy at 45%. The insurance sector presented a particularly interesting case, with 66% of its ads falling into the contextual band – often linked to life events within the conversation – rather than directly addressing the user’s immediate query. This suggests a reliance on association rather than direct relevance.

USER ENGAGEMENT AND THE VALUE OF CONTEXTUAL MATCHES
A key finding of the research is the prevalence of ads appearing in conversations lacking any purchase intent. A staggering 68% of ads were observed within conversations where the user exhibited no indication of wanting to buy, book, hire, or compare anything, regardless of the thread’s progression. This represents a significant waste of advertising spend, as these users are unlikely to be receptive to promotional material. The data further reveals that 40% of ad-carrying chats contained at least one irrelevant ad, and in 28% of instances, every single ad served was unrelated to the user's topic. This emphasizes the critical distinction between simply appearing in a conversation and achieving genuine engagement. OpenAI acknowledges this challenge, citing Denise Dresser's statement that ad dismissal rates have decreased by half since the advertising business launched. Dismissals are used as a proxy for relevance, indicating the system is improving. However, Searchable’s analysis, conducted roughly six months after the launch, suggests that the core issue of misaligned targeting persists, particularly given the platform's reliance on contextual hints rather than explicit keyword targeting. Ultimately, a placement within an assistant is not a substitute for being the direct answer, and ads that land in the wrong conversation are likely to be disregarded, while recommendations within a helpful response have proven to convert higher than organic channels.