A shopper browsing a retail app on a smartphone in front of a shop window with Black Friday sale signs

AI Shopping Assistants: Why Chatbots Still Struggle With Black Friday Deals

Picture the scene: a week before Black Friday, you ask a chatbot to find the best deal on a new television. It answers in seconds with a neat shortlist, prices and all. Then you click through, and the deal has vanished.

That’s roughly what happened when four popular AI assistants were put through a simple shopping test this month. The results were patchy, and they say a lot about how AI search treats the prices on your own website.

The tech firms are betting heavily on AI shopping. Amazon brought Alexa for Shopping to UK customers in November, OpenAI added a Shopping Research tool to ChatGPT just before last year’s Black Friday, and Google has started rolling out Universal Cart in the US, with price tracking and automatic promo code spotting.

A smartphone screen showing an AI chatbot app page against a pale, blurred background

One TV, Four Very Different Answers

The brief was the same each time: find a good deal on a 55-inch TV, using the free version of each tool. It wasn’t a scientific trial, but the gaps were hard to miss.

Go Ad Free

ChatGPT: bargains that weren’t there

ChatGPT’s favourite was a Hisense set on the Joybuy website at £499, reduced from £649. Its “safer” picks pointed to John Lewis and Richer Sounds at full price, yet neither retailer still listed the TV online. Another suggestion came in £150 over budget, and the answer finished with a £1,100 LG OLED marked as an advert.

Claude: brief and cautious

Claude gave the shortest reply, with little detail on specifications, and appeared to lean on a handful of online reviews. Two of its links went to dead product pages, others led to deal round-ups from Black Friday 2025, and it listed two Samsung models only to advise against buying them.

Gemini: a tidy table, no links

Gemini laid out four TVs in a comparison table but gave no links at all. One price was right, but the set was sold out at one of the shops it named. Another was correct at Argos, while Currys was charging £60 more than Gemini claimed.

Alexa: accurate, but only on Amazon

Alexa was the only assistant whose prices all checked out, from an Amazon Ember TV at £532 to a Samsung at £699, with a top pick cut from £749 to £616. The catch is obvious: it only searches Amazon. It did suggest checking other retailers for longer guarantees, although one of its four choices was a 50-inch model rather than the 55 inches asked for.

Agentic AI, where the assistant carries out several steps on its own, may do better. One tech journalist who tried ChatGPT’s agent mode during last year’s sales had a shortlist of seven TVs in about 20 minutes with very little effort. He did buy one of them in the end, but for less in a high-street shop, and he says he probably won’t use AI for deals this year because the live prices it returns are too often wrong.

AI Shopping by the Numbers

  • 19 countries were covered in a recent YouGov survey on AI and shopping, and the UK came second from bottom for the share of people who had used AI to help them buy.
  • 36% of respondents planning to shop over Black Friday had no intention of using an AI tool.
  • 1 in 4 assistants in the TV test returned prices that were all correct.
  • Half of chatbot answers about UK Black Friday tech deals mentioned Amazon, according to one analysis, with Currys close behind.
  • 7 out of 10 is the example consumer group Which? uses to describe the problem: ask for ten laptops under £400 and find seven of them cost more.
Grey hanging tags spelling SALE above the words Black Friday on a bright red background

What This Means If You Sell Online

For anyone running an online shop, the test is a warning. Chatbots don’t see the whole market. They keep returning the same few brands and the biggest retailers, and they often rely on whatever product information they last picked up.

Go Ad Free

Naureen Mohammed, who advises businesses on getting their products into AI results, says that when retailers don’t update prices as the sales begin, last year’s figures get presented as today’s deals. She also expects advertising to start shaping what AI tools recommend.

“You shouldn’t take any advice from the AI system at face value.”
– Naureen Mohammed

Which? senior researcher Hannah Walsh makes the same point from the shopper’s side. Wrong prices in AI answers add work rather than saving it, and they can also hide good products that really are within budget.

Shoppers will keep double-checking. Your job is to make sure that when an assistant does find your shop, what it finds is right.

Before the sales start, check that:

  • Sale prices are live on the product pages themselves, not just in a banner or a PDF.
  • Expired offers are removed or redirected, so old deal pages don’t keep circulating.
  • Sold-out products say so clearly, and discontinued ones point to a sensible alternative.
  • Your product structured data (price, currency and stock status) matches what’s on the page.
  • Shopping feeds, such as Google Merchant Center, update as often as your prices change.
  • Main specifications, like screen size, appear in the product title and description, not only in an image.

If a chatbot is going to recommend your products, give it the right numbers to repeat.

Go Ad Free

Create a free account, or log in.

Gain access to read this article, join developers who build better, faster, smarter solutions

Yes! I would like to receive new content and updates.

Scroll to Top