Yes, but only if you treat every price, deadline and stock claim an AI tool drafts as unverified until it matches your own records. AI will write "was $120, now $79, ends tonight" without knowing what you charged last month or whether the offer really ends. In the UK and the US, those claims are your responsibility, not the tool's. Check them against your records before anything goes out.
Sources checked 22 November 2026: the UK Competition and Markets Authority (CMA) and the US Code of Federal Regulations, each opened in the original. This is not legal advice, and I am not a lawyer.
Black Friday falls on 27 November this year and Cyber Monday on 30 November. For a small and medium-sized business (SMB, or SME in the UK and Europe) with an online shop, that means a week of offers drafted fast, often by AI, often by one person.
What did the CMA fine businesses for in October 2026?
The CMA fined three UK businesses a combined £6 million for misleading sales tactics, announced on 14 October 2026. Two of the cases are about exactly what an AI tool produces by default.
Wayfair was fined £4.8 million. According to the CMA, its "Daily Deals" and 24-hour "Flash Deals" used countdown timers to imply urgency, but many products stayed at the same or a lower price after the stated deadline. Appliances Direct was fined £1.2 million. The CMA found that pop-up banners advertised £10 to £25 discounts "for a limited time", yet the same discount codes recurred over many days. The third firm, Gold's Gym (UK) Limited, was fined £45,000 for leaving a joining fee out of the headline price.
All three admitted breaching the law and settled early, which earned a 40% reduction. The CMA also restated the ceiling: it can fine a business up to 10% of its global turnover, or £300,000 if that is higher. For a small business, the second figure is the one that matters.
None of these three was a tiny shop, but the rule is not size-dependent. The CMA's open letter on these claims is addressed to any "online business" selling to UK consumers. My reading, and it is an interpretation rather than something the CMA said about AI: AI makes this kind of breach cheap to commit, because offers can be drafted faster than anyone checks the facts inside them.
What counts as a price reduction claim or an urgency claim?
The CMA defines a price reduction claim as "any discount or special offer price that refers to a higher comparison price", and describes urgency claims as scarcity, popularity, "act fast" or time-limited claims, including countdown timers. Both definitions come from its open letter to online businesses, published 29 March 2023.
The letter's central test for the "was" price is that it "must be a genuine and realistic selling price", and that the business should be able to show it. Its examples of urgency problems include a promotion that does not end when it says it will, stock levels that are misrepresented, and timers that restart.
The CMA's guidance on unfair commercial practices, published 4 April 2025 and last updated 18 November 2025, covers conduct prohibited under the Digital Markets, Competition and Consumers Act 2024. That is the regime that gives the CMA its direct fining power, which it confirmed on 7 April 2025 was in force.
How do the UK and US approaches compare?
Both countries ask the same underlying question: was the higher price real? They differ in how specific they are about what "real" means. This comparison was checked at source on 22 November 2026.
| Question | UK (CMA) | US (16 CFR Part 233) |
|---|---|---|
| What must the "was" price be? | "A genuine and realistic selling price" (CMA letter, 29 March 2023) | "The actual, bona fide price at which the article was offered to the public" (section 233.1) |
| How long must it have been charged? | No number in the letter. The CMA's 1 August 2024 principles for online mattress sellers say compliance is more likely if the period is "at least 30 days" | "A reasonably substantial period of time", with no number given (section 233.1) |
| What about comparisons with other sellers' prices? | Covered by the same "genuine" test | Must be "based upon fact, and not be fictitious or misleading" (section 233.2) |
| Where is it written down? | Law plus CMA guidance | The Federal Trade Commission's Guides Against Deceptive Pricing |
The mattress principles are written for one sector and the CMA says they "do not introduce new law" and are a minimum standard, not a target. I treat them as the best published yardstick for anyone running a "was/now" promotion, because they are the only place I have found the regulator putting numbers on it. They also add a volume test: at least one product should have sold at the "was" price for every two sold at the discounted price.
The Ledger Test
I suggest one question for any offer an AI tool drafts. It is my own way of thinking about the problem, not established research.
The Ledger Test is the question I put to any AI-drafted offer before it goes out: could you show, from your own records, that every price, deadline and stock claim in it is true today?
If the answer is "I think so", the answer is no. A "was" price needs an order export or a price-change log behind it. A deadline needs an expiry that exists in your shop system, not only in the copy. A stock count needs to come from your stock system, not from a sentence the tool wrote.
What does a safe offer check look like?
Here is a one-page check you can copy. Run it on every offer, whoever or whatever drafted it.
The Offer Check
- Price. What was the "was" price? On which dates was it charged, on which website, and how many units sold at it? Attach the evidence. If you cannot, delete the "was" price and write only "now".
- Clock. Does the deadline exist in your systems as an expiry or a scheduled price change? What happens at the deadline, and who switches the offer off? If the same offer returns within days, do not call it "limited time".
- Stock. Is any scarcity message ("only 3 left") generated from live stock? If a person typed it, delete it.
- Total. Does the headline price include every mandatory fee, and are any extras added by a pre-ticked box? Delete the pre-tick.
- Sign-off. One named person, one date, one line saying which records they checked.
Take a hypothetical 12-person online homeware shop. It sold a lamp at £95 for six months, raised the price to £120 on 10 November, and plans a Black Friday banner reading "Was £120, now £95". The lamp has been at £120 for ten days. On the CMA's mattress yardstick that "was" price has not been established, and the banner would fail check 1. The honest options are "now £95", or a "was" price that had genuinely been charged and sold since 28 October or earlier. This is a hypothetical example, not a client case.
The Mistake I See Most Often
Across the business teams I train, nobody sets out to mislead a customer. The mistake is that the person approving AI-written copy reads it for tone and typos, not for truth. The wording sounds right, so it passes.
That is where I use the Loop Test. The Loop Test is three questions I ask before accepting "a human checks it" as an answer: Would that person notice a wrong result without being told to look for it? Would they notice in time to stop it mattering? And if they missed it, could the work be put back? Oversight that survives all three is a safeguard. Oversight that fails any of them is a signature.
Apply it to a Black Friday email. Would the approver notice a false "was" price without the price log open? Probably not. Would they notice before the email reaches the list? Only if they check before sending. Could it be put back? An email cannot be recalled, which is why the check has to come before sending. From 500+ keynotes in 30 countries, the lesson I keep returning to is that speed pushes checking to the end of the process, and the end rarely comes.
I set out the Receipt Rule in my piece on AI for sales and marketing in a small team, and about the seasonal rush in AI for retail and hospitality. If your shop uses AI to answer customers, the same discipline applies to promises made in chat, which I cover in AI for customer service in a small business.
What should you do this week?
- Pull the price history for every product in your Black Friday and Cyber Monday email, from your shop platform or order export.
- Delete every "was" price you cannot evidence, or that has not been charged for a sustained period. Use "now" prices only.
- Move the clock into your system. Set expiry on discount codes and scheduled price changes. Never let the copy promise an end time the system does not enforce.
- Give your AI tool the facts, and fence it in. Paste the verified price and the real end time into the brief, and add this line: "Use only the prices, dates and stock figures I have given you. Do not add urgency, scarcity, 'was' prices or 'last chance' wording that is not in this brief."
- Name one sign-off person and have them run the Offer Check, with the records open, on every offer.
- Keep a one-page record of each campaign: the offer, the evidence, who signed it off and when. If a regulator or a customer ever asks, that page is your answer.
If your team is using AI to write offers and nobody has shown them how to check what it produces, that is what my AI training for teams is built for. To see where your business stands before the next campaign, take the free Business AI Readiness Check, or talk to me about your team.
Sources and further reading
- Competition and Markets Authority, CMA fines firms £6 million for misleading sales tactics, GOV.UK, 14 October 2026.
- Competition and Markets Authority, Using urgency and price reduction claims online (open letter to online businesses), GOV.UK, 29 March 2023.
- Competition and Markets Authority, Discount and reference pricing principles: selling mattresses online, GOV.UK, 1 August 2024.
- Competition and Markets Authority, Unfair commercial practices, GOV.UK, published 4 April 2025, updated 18 November 2025.
- Competition and Markets Authority, CMA to boost consumer and business confidence as new consumer protection regime comes into force, GOV.UK, 7 April 2025.
- US Government, 16 CFR Part 233: Guides Against Deceptive Pricing, Electronic Code of Federal Regulations, read 22 November 2026.
Dan Fitzpatrick is a Forbes contributor, international keynote speaker and bestselling author who helps businesses adopt AI well. About Dan.



