AI-generated reviews are customer reviews written by a language model instead of a person who actually bought the product or used the service. Some are outright fakes posted under invented identities. Others are attack reviews aimed at a competitor. A growing share sit in a gray zone, where a real customer asks ChatGPT to tidy up what they meant to say.
All three land on the same star rating. And all three are changing what businesses should expect from the firms they hire to protect their reputation.
Five years ago, a reputation manager’s review work was mostly arithmetic. Get more happy customers to post, respond to the angry ones, nudge the average from 3.9 to 4.4. That model assumed a review was human-written evidence. It no longer can.
What counts as an AI-generated review?
Not every review touched by AI is fake. That distinction matters legally and practically, and plenty of business owners get it wrong in both directions.
A customer who ate at your restaurant, had a good meal, and asked an AI tool to fix their grammar has written a real review. The experience happened. The opinion is theirs.
A review is fraudulent when the experience or the reviewer is invented. A model writing twenty-five five-star posts for a dentist nobody visited. A competitor generating one-star complaints about a contractor who never worked for them. A business owner feeding an AI tool a list of “talking points” and posting the output under staff names.
The test isn’t whether software helped write it. The test is whether a real person had the experience being described.
How common are AI-generated reviews now?
More common than most business owners assume, and harder to measure than anyone would like.
Pangram Labs analyzed 30,000 front-page Amazon reviews in 2025 and flagged 3% as AI-generated. Nearly three-quarters of those were five stars, and 93% carried the “Verified Purchase” badge, the label shoppers lean on as proof a review is real, according to reporting in Inc.
Three percent of front-page reviews is a lot. Those are the reviews that actually move purchase decisions.
The broader fake review problem is bigger still. Capital One research published in 2026 estimated that about 30% of online reviews are “fake or ungenuine” and that fake reviews are growing 12.1% faster than reviews overall. The same research found that 82% of consumers encountered a fake review at least once in the past year.
Platforms are removing them in staggering numbers. Trustpilot took down 4.5 million fake reviews in 2024, roughly 7% of everything submitted. Google says it blocked 292 million policy-violating reviews on Maps in 2025.
Those removal figures are the floor, not the ceiling. They only count what got caught.
Why fluent fakes broke the old playbook
For years, spotting a fake review came down to reading it. Fakes were short and vague. They overused words like “amazing” and “best ever.” They showed up in clumps from accounts with no history.
Language models erased the first half of that checklist overnight. A fabricated review can now run 150 words, name a server, complain mildly about parking, and still land on five stars. It reads like a regular customer on a good night.
What still exposes fakes is the data around the words. Account age. Device and IP overlap. Review velocity against a business’s normal baseline. Whether the reviewer’s location history puts them anywhere near the business. Trustpilot says about 90% of its fake review removals now happen automatically through machine learning, and Google uses its Gemini models to catch violating content before it publishes.
So the fight has moved. It’s AI writing reviews against AI screening them, with metadata as the referee. A reputation team that still evaluates reviews one at a time is bringing a magnifying glass to a server farm.
The FTC already decided how this works
Federal law caught up faster than most people expected.
The FTC’s Rule on the Use of Consumer Reviews and Testimonials took effect October 21, 2024. It bans reviews that misrepresent themselves as coming from a real person or reflecting a real experience, and the Commission’s announcement cited AI-generated fake reviews as an example. It also prohibits buying reviews, paying for reviews that express a particular sentiment, posting undisclosed reviews from employees or executives, and suppressing negative reviews through threats.
The penalty structure is what should get attention. Violations can cost up to $53,088 each, according to a 2026 analysis by law firm Benesch published on Mondaq. Each fake review is a separate violation.
Run that against a model that can produce 500 reviews in a minute, and you see the problem. A shortcut that costs pennies to generate can carry a theoretical liability in the tens of millions.
Enforcement is real, too. The FTC sent warning letters to roughly a dozen companies in December 2025. In April 2026, it settled with TruHeight over fake employee and incentivized reviews, with a $4 million judgment. A month later, the FTC and the Illinois Attorney General sued Premium Home Service over fake business listings padded with fabricated five-star reviews, per DLA Piper’s review of 2026 enforcement.
AI summaries made every review louder
Here’s the part that most coverage misses, and it’s the part that matters most for reputation work.
Shoppers increasingly don’t read reviews. They read what an AI says about the reviews. Google now generates review summaries in Maps and Search. ChatGPT, Gemini, and Perplexity answer “is this business any good?” by synthesizing review text, articles, and forum threads into a paragraph.
BrightLocal’s 2026 consumer survey found 45% of consumers have used tools like ChatGPT to get local business recommendations. Omnisend found 63% of U.S. consumers use AI while shopping.
A language model summarizing reviews pulls out themes. If thirty fabricated one-star reviews all mention “billing surprises,” that phrase can surface in an AI answer about your business. It can stay there after some of the fakes come down, because the summary was already generated, cached, and repeated across other sources.
One fake review used to cost you a fraction of a star. A coordinated batch can now write the sentence an AI assistant uses to describe you to every new prospect.
That shift is why review management and AI visibility are no longer separate jobs.
What AI-generated reviews change for online reputation management companies
The work of online reputation management companies used to be judged by output: reviews gained, average rating lifted, negative links pushed down. Those numbers still matter. They just can’t be the whole pitch anymore, because the cheapest way to hit them is now the riskiest.
Here’s what the job looks like when it’s done right in 2026.
- Provenance-first review generation: Every review request should trace back to a real transaction. That means invitations triggered from CRM or point-of-sale records, sent to actual customers, with no strings tied to sentiment. If a platform or regulator asks where a review came from, the answer should be a timestamped record.
- Baseline and anomaly monitoring: A good firm knows a client’s normal review rhythm. Eight reviews a month that suddenly jump to forty in a weekend are worth investigating, whether the new reviews are glowing or brutal. Positive spikes can trigger platform filters too.
- Forensic removal requests: Platforms remove reviews that break their policies, not reviews that are merely unfair. Strong removal requests document the pattern: account creation dates, no matching customer record, timing clusters, recycled details across posts. Fluent text won’t give a fake away, so the case has to be built from metadata.
- Responses written for two audiences: A public reply is read by the prospect and ingested by the AI summarizer. A calm, specific response noting that no record of the visit exists protects you in both places. Silence lets the fake review’s framing stand unchallenged.
- AI answer monitoring: Reputation now lives partly inside ChatGPT, Gemini, and Google’s AI Overviews. Firms need to check what those tools say about a client, which sources they’re citing, and whether a review attack has bled into the answer.
- Broader content footprint: AI tools don’t rely on reviews alone. They pull from news coverage, profiles, and third-party mentions. Credible, independent content gives the models something better to quote than a cluster of suspicious one-star posts.
NetReputation, for example, pairs business review management with fake review removal and AI visibility monitoring because a review problem fixed only on the review platform isn’t fully fixed.
How do you choose between online reputation management companies now?
The industry still has firms selling the old playbook with fresh branding. A few questions separate them quickly.
- Have you ever written, generated, or posted a review on a client’s behalf? The only acceptable answer is no. Any hesitation tells you what you need to know.
- How do you get new reviews? Listen for real customer outreach tied to real transactions. Walk away from “review packages,” “seeded profiles,” or a guaranteed number of new five-star posts per month.
- What do you do when a client gets hit by an AI-generated attack? A competent firm will discuss metadata, platform escalation paths, documentation, and when to involve legal counsel. Vague talk of “burying” the reviews is a weak answer.
- Can you show me what AI assistants currently say about my business? Firms that are paying attention can run this check in the first conversation.
- How do you price the work? Be cautious with flat promises like “50 reviews for $X.” Legitimate reputation work depends on the situation, the platforms involved, and how much damage is already baked into search and AI results.
Any company that guarantees a star rating is indirectly telling you how it plans to get there.
What businesses should do before the next review attack
You don’t need to wait for a crisis to prepare. A short checklist covers most of the exposure.
- Audit your review sources. Make sure nobody on staff, at an agency, or at a franchise location is using AI to write or “seed” reviews. Under the FTC rule, the business can be liable for what its agents do.
- Connect review requests to real transactions so you can prove provenance later.
- Record your normal review velocity per platform. You can’t spot a spike without a baseline.
- Ask ChatGPT, Gemini, and Perplexity what they know about your business, and save the answers. Repeat monthly.
- Keep a simple incident file: screenshots, dates, reviewer profile links. Removal teams and attorneys will ask for it.
The businesses that get through review attacks fastest are the ones that already have records.
The trust premium goes to whoever can prove it
Generative AI made fake reviews almost free to produce and very hard to catch by eye. Platforms answered with their own models. Regulators answered with a rule that names the technology and prices each violation at five figures.
That squeezes online reputation management companies toward one defensible offer: authenticity with receipts. Every review traceable to a real customer. Every removal request backed by evidence. Every response written for the person reading it and the model summarizing it.
The firms still selling volume will keep getting cheaper. Clients will keep paying for them twice.
FAQ
What are AI-generated reviews? AI-generated reviews are reviews written by a language model rather than by a customer describing their own experience. When the reviewer or the experience is invented, they are fake reviews. When a real customer uses AI only to polish their own honest opinion, the review is still legitimate.
Are AI-generated fake reviews illegal? Yes. The FTC’s Consumer Review Rule, effective October 21, 2024, bans reviews that misrepresent a real reviewer or real experience and names AI-generated fakes as an example. Each violation can carry a civil penalty of up to $53,088, and businesses can be liable for reviews posted by employees or agencies.
How can you tell if a review was written by AI? Reading the text is no longer reliable because language models write fluent, detailed reviews. Stronger signals come from metadata: new or empty reviewer accounts, review spikes far above a business’s normal pace, overlapping devices or IP addresses, and reviewers with no plausible connection to the business’s location.
What do online reputation management companies do about AI-generated reviews? Reputable online reputation management companies monitor review velocity for anomalies, build evidence-based removal requests from reviewer metadata, write public responses, track how AI assistants describe the client, and generate new reviews only from verified customers. They escalate to legal counsel when an attack looks coordinated or extortionate.
Can an online reputation management company write reviews for my business? No legitimate firm will. Writing or generating reviews for a client violates the FTC rule and every major platform’s policies. A reputation firm can legally run review request programs that invite real customers to share honest feedback, with no conditions on whether the review is positive.
Do fake reviews affect what ChatGPT and Google AI say about a business? They can. AI summaries in Google and assistants like ChatGPT draw themes from review text, so a batch of fake reviews repeating the same complaint can shape how a business is described. That framing can persist after some reviews are removed, which is why AI answer monitoring now belongs in review management.
