← AI, commerce and societyA6 · LESSON 4 OF 7

A6, Lesson 4: Fake reviews, fake listings, deepfake endorsements

Tags: Sourced stories / Our view / Illustrative examples. Self-paced reading and practice. No math or code. Every case or news claim carries its source and the date we read it; invented figures are labeled invented.

1. The story

In February 2024, a finance worker in Hong Kong joined a video call with what looked like his company's chief financial officer and several colleagues. Every face on the call was a deepfake. Convinced, he transferred 25 million dollars to the scammers, reported CNN. Months later, on August 14, 2024, the US Federal Trade Commission announced a final rule - 16 CFR Part 465, in force since October 21, 2024 - against fake or false reviews and testimonials: businesses may not create or sell reviews that materially misrepresent whether the reviewer exists or what they experienced, and buying or spreading such reviews is banned when the business knew or should have known. AI-generated fakes are one named example, and knowing violations can bring civil penalties.

2. The one idea

AI made trust cheap to fake, so proof gets more valuable. A glowing review, a polished listing, even a familiar face on a video call once took real effort to fake. Now they cost little - so the honest version is worth defending, and the fake version is worth planning for.

3. The kitchen-table version

A queue outside a restaurant used to be proof you could not buy. Hiring fifty people to stand in line is cheap, so the queue stops being proof and becomes a claim. Your job as a buyer is to look for what is still expensive to fake; your job as a seller is to keep producing it.

4. Worked example

Our invented candle shop has collected 86 reviews across three years - about 2.4 a month, each one slow and real. A competitor listing appears with 40 glowing five-star reviews in its first week, all in the same cheerful rhythm. That week-one burst is roughly seventeen times the shop's monthly pace (40 against 2.4). Figures invented and script-checked. Pace and sameness are signals, not proof: maybe the competitor shipped a viral batch, maybe not. The honest shop cannot control the fake listing; it can control whether its own proof - specific replies to unhappy customers, answered questions, photos from real buyers - is easy to find.

Instructor: "So report the listing and wait?" Learner: "Report it, sure. But the durable play is making your own proof dense and visible, because a buyer who cannot tell real from fake will default to whoever looks most alive."

5. Why this matters to a shop

Your real reviews are both an asset and a target. In the United States, the FTC's Rule on the Use of Consumer Reviews and Testimonials (16 CFR Part 465, in force since October 21, 2024) bans businesses from creating or selling reviews that materially misrepresent whether the reviewer exists, really used the product, or what happened - and bans buying or spreading such reviews when the business knew or should have known. Courts can impose civil penalties for knowing violations. Note what it does not do: it does not ban AI assistance or paid reviews in general; it targets misrepresentation, insider reviews without disclosure, and incentives conditioned on a particular sentiment. And your own team is a target on the other side: a video call or a familiar voice is no longer proof of a person.

6. What it costs and what it breaks

Collecting genuine reviews is slow; faking them is fast, which is exactly why the rule and the platforms punish it. Second-channel checks add minutes to an urgent payment - that friction is the point, and it is far cheaper than one Hong Kong-style mistake. There is no reliable way to spot every fake by reading; plan for a world where you cannot always tell.

7. Our habits, not rules

Never buy reviews, and make your real proof easy to find - answered questions, specific replies to unhappy customers, responsive service. Move money only after a second-channel check: if a "boss" or "supplier" asks for payment on a call, confirm on a channel you already trust, like the saved phone number in your records.

8. Monday checklist

List where your reviews live and who can answer them. Check your own practices against the rule: no incentives tied to positive sentiment, and any insider review disclosed. Write your second-channel payment rule on one line and say it to the team out loud. Decide who watches your category for suspicious new listings. Pick one piece of real proof - a customer photo, a detailed reply - and make it more visible this week.

9. Try it yourself (five minutes)

Take ten reviews from any listing, yours or a competitor's, and try to sort them into real and generated. Then sit with how unsure you are. That uncertainty is the lesson - and it is why your own visible signs of life matter more each year.

10. Self-check

  1. What does the FTC rule ban outright, and what only with knowledge? Answer: Creating or selling fake or false reviews is banned outright; buying or spreading them is banned when the business knew or should have known.
  2. Does it ban paid or AI-assisted reviews in general? Answer: No - it targets misrepresentation, undisclosed insider reviews, and incentives conditioned on a particular sentiment.
  3. What made the Hong Kong deepfake call work? Answer: Every face was fake and the request stayed on one channel; a single known-number callback would have broken it.
  4. Why is "I can spot a fake review" a dangerous confidence? Answer: Because there is no reliable way to tell by reading, so confidence replaces process.
  5. What is your second channel for payment requests, and does your team know it?

Sources

  • CNN, "Finance worker pays out $25 million after video call with deepfake 'chief financial officer'," February 2024 ($25M; about HK$200M per Hong Kong police), read 3 Nov 2026.
  • FTC, "Rule on the Use of Consumer Reviews and Testimonials" (16 CFR Part 465), announced August 14, 2024, in force since October 21, 2024: businesses may not create or sell reviews that materially misrepresent the reviewer's existence or experience, or buy or spread them when they knew or should have known; civil penalties for knowing violations; AI-generated fakes are one named example; it does not ban AI assistance or paid reviews as such.
  • Rule structure checked against the FTC press release and the eCFR rule text, 3 Nov 2026.
  • The review-pace figures (86 over three years, 40 in a week, 2.4/month, seventeenfold) are invented, illustrative and script-checked.
  • Review flags: platform rules are mentioned generically and not quoted; enforcement actions beyond the rule's effective date are not claimed.
  • Go deeper, free: Harvard's session 11, Misinformation, covers mis- and disinformation and the "liar's dividend"; its slide deck adds the Hong Kong deepfake fraud and the falling cost of scam email. Watch it here.

Review flags

  • platform rules are mentioned generically and not quoted; enforcement actions beyond the rule's effective date are not claimed.

Claim ledger

Claim or exampleBasisChecked
Hong Kong finance worker paid USD 25 million after a deepfake video call.CNN, deepfake call; FTC rule announcement3 Nov 2026
US rule: creation/sale scope differs from purchase/dissemination knowledge standard.CNN, deepfake call; FTC rule announcement3 Nov 2026

Independence

Independently developed by Commerce Stories, not affiliated with Harvard. We link to the archived 2024 course; we do not reproduce its videos, slides or exercises. The linked videos were not watched for this written-material review.