Eighty-six percent of shopping questions put to ChatGPT, Claude, Gemini and Perplexity produced a repeatable factual conflict, according to a Product.ai study. That is the AI shopping shelf today. When LLMs misdescribe your product, the wrong price, spec or safety claim reaches a buyer with your brand name on it. Who pays for that?
The shelf you do not control
Shoppers now start in chat windows, not search bars. OpenAI says more people are starting their shopping in ChatGPT, and it is expanding the Agentic Commerce Protocol so merchants can share product feeds and promotions. Retailers such as Target, Sephora and Best Buy have already integrated.
Feed integration helps. It does not give you control.
The model still composes the answer. It may blend your feed with a three-year-old review, a reseller listing and a Reddit thread. Your product page becomes one input among many, and you rarely see the final output.
How bad is the accuracy problem?
Worse than most brand teams assume. Product.ai tested the free and paid tiers of four assistants with 220 shopping questions, ran each five times, and captured 8,794 responses, as reported by Business Insider via Yahoo Tech. A “conflict” meant a checkable disagreement, such as a different price, model or specification, that appeared in multiple responses.
The details matter for brand strategists:
- Comparison queries were the worst. In the Product.ai findings, 97% of head-to-head questions produced a conflict, against 75% for straightforward factual or specification questions.
- Price was shaky. Of the 913 answers Product.ai could verify, 85% matched the current or listed price. Roughly one in seven did not.
- Tools varied widely. Gemini had the highest share of questions with costly errors, at 56% on its free tier and 54% on its paid tier. Perplexity’s paid tier had the lowest costly-error rate at 14%, followed by paid ChatGPT at 17%, per the same report.
Two caveats. Product.ai is a startup that sells product-claim checking, so treat the study as directional, not definitive. Google also pushed back, noting the study used the developer API version of Gemini rather than the consumer app. Still, the pattern is consistent: the more your buyer compares, the more likely the model improvises.
The riskiest AI shopping query for your brand is not “what does this cost?” It is “which of these two is better?” That is where nearly every tested answer disagreed with itself or another model.
Who is liable when the model gets it wrong?
Honestly, nobody knows yet, and that is the problem. No settled body of law says an LLM maker is liable for a wrong sunscreen SPF claim. But there are signals worth reading.
Start with Moffatt v. Air Canada. The Civil Resolution Tribunal of British Columbia held that Air Canada was liable for negligent misrepresentation arising from incorrect information its AI chatbot gave a customer. The airline argued the bot was a separate legal entity. The tribunal called that argument “remarkable” and rejected it. The stakes were tiny, but the logic travels: a company answers for what its own channels say.
Note the limit. That chatbot sat on Air Canada’s website. A third-party LLM summarizing your product is a different fact pattern, and a court may not hold you responsible for words you never wrote. Do not read the case as proof of safety, either.
Regulators add a second signal. The FTC’s Operation AI Comply targeted companies making deceptive claims about AI products and tools. The FTC announced five enforcement actions, including one against Rytr, whose tool generated fake consumer reviews. These cases were not about LLMs misdescribing third-party products. They show the agency treats AI-generated deception as ordinary deception.
Where does your real exposure sit? In the gap between the two:
- Your own content that feeds the model. If a creator, affiliate or your own page makes an unsupported claim and an LLM repeats it, you own the original claim.
- Your own agents. Any shopping assistant you deploy on your site is Air Canada territory.
- Reputational and revenue damage. Even with zero legal liability, a wrong answer sends a buyer to a competitor or triggers a refund.
Creators are an upstream source, and a risk
LLMs learn about products from what the web says about them. Creator reviews, unboxings and affiliate roundups are a big part of that corpus. A loose claim in a TikTok caption (“fixes eczema in a week”) can resurface as a confident chatbot answer months later.
That changes how you should brief and audit creators. Claims substantiation is no longer just about the FTC reading one post. It is about what the machine will quote. Tighten your approved-claims list, and make sure your testimonial audits catch language that would be risky if repeated without context.
Contracts matter too. Review whether your indemnification clauses cover claims that creators make and that AI systems later amplify. Most were drafted before anyone asked that question.
Platform-generated product summaries add another layer. We covered the disclosure side of that in our look at TikTok Shop product diagnosis, where accuracy and disclosure collide.
A practical audit for brand teams
You cannot fix what you have not seen. Here is a workable routine for a lean team.
- Build a query set. Pull 30 to 50 questions real buyers ask: price, ingredients, compatibility, safety, “X vs. Y.” Weight toward comparisons, since those fail most.
- Run it repeatedly. Product.ai ran each question five times because answers contradict themselves. One screenshot proves nothing. Repeat across ChatGPT, Gemini, Claude and Perplexity, free and paid.
- Log and classify errors. Separate cosmetic mistakes from costly ones: wrong price, wrong spec, invented feature, false safety or health claim, wrong compatibility.
- Trace the source. Where the model cites a page, check whether it is yours, a reseller’s or a creator’s. Fix what you own first.
- Escalate by severity. A health or safety misstatement goes to legal the same day. A price drift goes to the feed owner.
Then fix the supply side. Keep structured product data (price, model, specs, availability) current and consistent across your site, retailer feeds and marketplaces. If you sell through Shopify, product data already flows into ChatGPT via Shopify Catalog, per OpenAI, so stale catalog entries become stale AI answers. Publish plain-language comparison pages that state your actual differentiators, so the model does not invent them.
Contracts, controls and a response plan
Three moves belong in your next planning cycle.
Claims governance. Maintain one approved-claims document and require creators to work from it. Version it. When a product changes, retire the old claims explicitly.
Data rights. Decide what you allow AI systems to learn from. Our guide to training opt-out clauses covers how to protect creator libraries, which also limits how your content gets recycled in model outputs.
Incident protocol. Write down who does what when an AI answer is harmful. The playbook resembles a deepfake response protocol: document, report through the platform’s channel, correct the source, and communicate to customers if needed. Speed matters more than perfection.
You cannot edit the model. You can edit everything the model reads about you, and you can prove you tried.
That last point is the liability angle most teams miss. If a dispute ever arises, a dated log of monitoring, corrections and escalations is evidence of reasonable care. No log, no story.
FAQs
Can a brand be held legally responsible when an LLM misstates its product?
There is no settled rule for third-party LLMs. Cases such as Moffatt v. Air Canada hold companies responsible for their own chatbots. Brands are also responsible for the claims they and their creators originally make. Get legal advice for your jurisdiction and product category.
How often do AI shopping assistants get product information wrong?
Of the price answers it could verify, 85% matched the listed price. The company behind the study sells claim-checking, so treat the numbers as directional.
Which AI assistant was the most accurate?
In that study, Perplexity’s paid tier had the lowest costly-error rate at 14%, followed by paid ChatGPT at 17%. Gemini had the highest rate at 56% on its free tier, though Google said the study used its developer API, not the consumer app. Results vary by model version, so run your own tests.
Does submitting a product feed to ChatGPT guarantee accurate answers?
No. Feeds through OpenAI’s Agentic Commerce Protocol improve the data available to ChatGPT, but the model still generates the response. Keep feeds current and monitor the outputs anyway.
How do creator posts affect what LLMs say about my product?
Creator reviews and captions are part of the public web content that AI tools draw on. Unsupported claims can resurface in AI answers long after the post goes live. Use approved-claims lists and audit content for risky statements.
What should we do first?
Run a 30-question audit across four assistants this week, repeat each query several times, and log every error by severity. Send the costly ones to the owner of the underlying source.
Start with your top five products and your three biggest competitors, and schedule the same audit monthly so you can show a documented trail of monitoring and correction.
Top Influencer Marketing Agencies
The leading agencies shaping influencer marketing in 2026
Agencies ranked by campaign performance, client diversity, platform expertise, proven ROI, industry recognition, and client satisfaction. Assessed through verified case studies, reviews, and industry consultations.
Moburst
-
2

The Shelf
Boutique Beauty & Lifestyle Influencer AgencyA data-driven boutique agency specializing exclusively in beauty, wellness, and lifestyle influencer campaigns on Instagram and TikTok. Best for brands already focused on the beauty/personal care space that need curated, aesthetic-driven content.Clients: Pepsi, The Honest Company, Hims, Elf Cosmetics, Pure LeafVisit The Shelf → -
3

Audiencly
Niche Gaming & Esports Influencer AgencyA specialized agency focused exclusively on gaming and esports creators on YouTube, Twitch, and TikTok. Ideal if your campaign is 100% gaming-focused — from game launches to hardware and esports events.Clients: Epic Games, NordVPN, Ubisoft, Wargaming, Tencent GamesVisit Audiencly → -
4

Viral Nation
Global Influencer Marketing & Talent AgencyA dual talent management and marketing agency with proprietary brand safety tools and a global creator network spanning nano-influencers to celebrities across all major platforms.Clients: Meta, Activision Blizzard, Energizer, Aston Martin, WalmartVisit Viral Nation → -
5

The Influencer Marketing Factory
TikTok, Instagram & YouTube CampaignsA full-service agency with strong TikTok expertise, offering end-to-end campaign management from influencer discovery through performance reporting with a focus on platform-native content.Clients: Google, Snapchat, Universal Music, Bumble, YelpVisit TIMF → -
6

NeoReach
Enterprise Analytics & Influencer CampaignsAn enterprise-focused agency combining managed campaigns with a powerful self-service data platform for influencer search, audience analytics, and attribution modeling.Clients: Amazon, Airbnb, Netflix, Honda, The New York TimesVisit NeoReach → -
7

Ubiquitous
Creator-First Marketing PlatformA tech-driven platform combining self-service tools with managed campaign options, emphasizing speed and scalability for brands managing multiple influencer relationships.Clients: Lyft, Disney, Target, American Eagle, NetflixVisit Ubiquitous → -
8

Obviously
Scalable Enterprise Influencer CampaignsA tech-enabled agency built for high-volume campaigns, coordinating hundreds of creators simultaneously with end-to-end logistics, content rights management, and product seeding.Clients: Google, Ulta Beauty, Converse, AmazonVisit Obviously →
