You spent the year putting your brand in front of AI shoppers. You optimized your product feeds for ChatGPT. You wrote the comparison pages the language models would pick up. You paid the GEO vendor to keep your citations fresh. Last week, Semrush surveyed 2,338 U.S. adults and found that nearly 81 percent of consumers who use chatbots for online shopping have decided against a purchase based on AI guidance. The same survey found 74 percent of all consumers would be at least somewhat less likely to buy a product if a chatbot surfaced mixed or negative reviews about it. The channel you optimized for is now the channel most likely to take you off the consideration list.
This is not a story about Google losing search share to ChatGPT. Only 29 percent of all consumers use AI chatbots for product research today. It is a story about a new consideration gate, and most brands have not adjusted for it yet.
The new gate is the answer, not the ranking
The old assumption was that AI worked like a better search engine. If your product ranked in the top three results, the chatbot would mention it, the shopper would click through, and your conversion rate would look the way it always had. The Semrush data shows that assumption is wrong in two ways at once.
First, AI recommendations work in both directions. About 58 percent of AI users have bought something a chatbot recommended. About 58 percent have decided against a purchase based on information the chatbot provided. Those are the same people, not separate segments. The chatbot is not a one-way megaphone for your brand. It is a two-way filter, and the negative filter is at least as powerful as the positive one.
Second, the chatbot is the gate at a moment when the shopper is already skeptical. Among consumers who use AI at all, 71 percent believe chatbots can recommend the best brands and products. Among the same group, 85 percent say they would be less likely to buy if a chatbot surfaced mixed or negative reviews. The shopper trusts the AI enough to take its negative recommendation seriously, and they have a long list of complaints waiting to be surfaced.
Buying visibility inside the chatbot does not fix this
The natural response is to buy chatbot ad placements. ChatGPT, Perplexity, and Google are all starting to sell placements inside the answer. The math feels right. You get a slot in front of a high-intent shopper at the moment of decision. You can measure the click-through and the conversion.
The Semrush data says the math does not work. Forty-two percent of consumers say they dislike chatbot ads. Among those who dislike them, 66 percent say the ads make them question the integrity of AI responses. An ad in the answer trains the shopper to discount the answer, which discounts your placement, which discounts the conversion you were promised. You are paying to make the channel less effective for everyone, including yourself.
The harder problem is the one the ad does not touch. The shopper asks the chatbot, without you in the room, whether your product is worth the price, whether customers have had problems with it, and how it compares to the cheaper alternative. The answer the chatbot gives is shaped by reviews, Reddit threads, forum complaints, and the comparison content your competitors wrote. You are not at the table for that conversation. You cannot buy a seat at that table. The ad buys you a slot in a different conversation, and the second conversation is the one that decides whether the shopper buys from you at all.
What to actually do this quarter
Stop measuring how often AI mentions your brand. Start measuring what AI says about your brand when no one is looking. Pick five prompts a real shopper would type, run them against ChatGPT, Claude, Perplexity, and Gemini, and read the answers the way a skeptical buyer would. The pattern is usually the same. The chatbot knows your category. The chatbot does not know your product. The chatbot pulls the negative from wherever it can find it. If the answer makes a careful shopper hesitate, the answer is your real optimization target.
Audit your third-party review surface next. The chatbot is only as honest as the public material it can read. If your Amazon reviews are full of one-star complaints about a defect you fixed eighteen months ago, the chatbot does not know that. If your support forum has a long thread about a billing issue you resolved in May, the chatbot does not know that either. The information architecture you built for humans is now an information architecture for agents. Treat it like SEO in 2009, before anyone was watching.
Finally, build an answer layer for the questions you cannot afford to lose. Write the comparison content. Publish the customer story that explains the negative review. Make the price justification explicit. The chatbot does not need to be persuaded. The shopper does, and the chatbot is the only thing standing between the two of them. Every unaddressed objection becomes an AI-generated reason to buy from someone else.
Sources
MarTech, AI is telling consumers not to buy your product, September 18 2026. Semrush AI Chatbots Talk AI Users Out of Buying survey, September 2026, n=2,338 U.S. adults.