A study by IP protection firm Podqi, shared exclusively with The Verge on 12 November 2026, tested ChatGPT, Gemini and Perplexity with a handful of shopping questions. It found that shoppers who use these chatbots to research products or find places to buy could be directed to imposters and other risky websites.

I could not load the full article, so I am working from the summary only. I won’t quote numbers or examples I haven’t seen.

The excitement is about agents that shop for you

The pitch for AI shopping is simple. You ask the chatbot what to buy and where, and it answers. Soon an agent does the buying as well. Fans of this treat the chatbot as a trusted salesperson.

I start with a different question. What actually changed? Shoppers always landed on bad sites. Search results had fake shops, paid ads had fake shops, and social feeds had them too. What changed is that the answer now arrives as one confident sentence, with no list of ten links to look over. The shopper has less to judge by.

What didn’t change is that the chatbot can only work from data it can read. I can’t tell from the summary why these answers went wrong. My view is that when a real seller has not stated its facts clearly anywhere an agent looks, the gap gets filled by whatever else is there. Imposters are happy to fill it.

For a small seller, the shift is already decided

Agentic commerce is still too early to call. Big platforms may open up to agents or block them, and the giants can wait to see how it turns out. A small business can’t wait.

For smaller businesses the direction is already settled. Get data-centric. Your website is one place your data shows up, and agents read many others. Wherever your data appears, it needs to be precise, detailed and accurate about what you are.

Think of a small online shop selling handmade candles. A customer asks a chatbot where to buy from it. If the shop’s official domain, payment methods, delivery terms and product details sit only inside a few designed pages, the chatbot has little to quote. A copycat site with the same product photos and a similar name gives it just as much to quote.

Publish your own facts, in plain numbers

The shop’s answer is to state its canonical facts and put them where agents read:

  • The official domain, spelled out, and which domains are not yours.
  • Payment methods, and which ones you never use.
  • Delivery details: where you ship, how long it takes, who carries it.
  • Specific product attributes. Sizes, materials, burn times, ingredients.

Engines quote specific facts. Marketing adjectives get dropped. “Premium, artisan, loved by thousands” gives an agent nothing to repeat. “Soy wax, 40-hour burn, ships from Singapore in two working days” does. Use your own numbers, of course.

This also has to reach beyond your own site. Your own pages get you cited fast. Other people’s pages keep you there. Marketplace listings, directories and review sites should all say the same thing about you.

Check what the chatbots say about you

I trust nothing unverified, and that includes what a chatbot says about my own business. So the first step costs nothing. Ask ChatGPT, Gemini and Perplexity the questions a customer would ask:

  • Where can I buy from your brand?
  • What is your official website?
  • Do you deliver to my area?
  • What payment methods do you accept?

Write down what each one says and the date. AI answers change week to week, so a result from last month proves little. If a chatbot names a wrong domain, a wrong delivery promise or a shop that isn’t yours, you have found the gap. Fix the facts at the source, then ask again.

Where this ties to my own work

When I worked on AeroChat’s AI search visibility, we wrote down exactly what AeroChat is, who it’s for and how it’s different, with no marketing language. Then we built pages around the questions that produced an AI answer. Within three weeks, AeroChat was in the same Google AI Overview answer box as Tidio, Gorgias and Intercom. That was one product, on Google’s AI Overviews, in early 2026. It is not a promise for your shop. It does show that clear facts get read.

What to do this week

  1. Run the four questions above on three chatbots and record the answers with the date.
  2. Write one page that states your official domain, payment methods, delivery terms and product attributes in plain numbers.
  3. Make your marketplace listings, directory entries and social profiles match that page.
  4. Repeat the check every month and correct whatever the chatbots get wrong.

Source: The Verge AI: AI chatbots are recommending fake retail websites, study finds