In 1999, I co-founded Epinions, one of the first user-generated review sites, and helped pioneer the reviews economy. The premise was that anyone could review anything, and the collective judgment of the masses would sum to something useful.

Epinions succeeded. We sold the company to eBay. But it, along with the review sites that followed, set an industry in motion that would eventually be gamed. Today’s online reviews are a mix of the genuine and the fabricated: real experiences sitting next to anonymous ratings that can be manufactured or bought, often indistinguishable from the real thing. It got bad enough that in 2024, the Federal Trade Commission finalized a landmark rule making deceptive review and testimonial practices illegal.

I had a stake in that fight in 1999, and I find myself with one again now. I co-founded Nextdoor, a neighborhood network built on real identity and local trust. A large part of how neighbors use Nextdoor is to gather local advice and recommendations — reviews, in effect — on local businesses and service providers. My co-founders and I didn’t set out to get back into the reviews business. But here we are, and the timing works in our favor: in today’s world, who’s behind the review matters more than ever.

The platforms at the center of this problem aren’t oblivious to it. They’ve built out enforcement teams, sued fake-review brokers, and removed hundreds of millions of fraudulent reviews over the past year alone. The fake and manipulated share of reviews kept growing anyway. That’s not evidence the effort isn’t real. It’s evidence that policing anonymous content after the fact can’t keep pace with how cheap it’s gotten to fake it in the first place.

Eighty-five percent of consumers now suspect online reviews are fake, and eighty-eight percent say they can’t tell which ones are real. None of that has killed the habit. People still want reviews and still act on them. What’s eroding is their ability to tell which ones deserve their trust. What they actually want to know is who’s telling them this, and why they should believe them.

What makes a recommendation trustworthy?

What the latest issues with the review industry have shown us is that any trust signal is only as durable as what it costs to fake and what you lose when you’re caught. Star ratings fail both tests. There’s no way to know who’s behind the rating, or what their bar for five stars even is. An anonymous rating costs nothing to manufacture, and no one pays a price when it’s exposed.

An accountable written recommendation passes both tests. When a neighbor recommends a contractor under their real name, you know exactly who’s vouching for them, and what that neighbor stands to lose, in reputation and relationship, if the contractor wrecks your kitchen. The reviewer has skin in the game. The star never did.

Early research fielded this summer by the consumer-insights platform Echo points in the same direction: nearly half of respondents said they’d suspected a fake or manipulated review in the past year, and 81% said they’d trust a recommendation more if the person making it were held socially accountable.

When a machine answers instead of a person

Ask a model today to name the best pizza place in town and you’ll get a confident, instant answer for free. But abundance always changes what’s scarce. Information used to be scarce. Now it’s free and instant. Provenance isn’t.

The machines giving those answers are trained on the same reviews we already can’t trust. It turns out they’re no better than we are at telling which ones are real. A 2025 study out of Nottingham University Business School found that AI-generated fake reviews are now indistinguishable from real ones, to human readers and to other AI models alike. This reality just creates more unreliable confusion for the end consumer.  

And AI presents problems for the businesses being reviewed. They have no way to influence the answers the LLMs produce, and even more problematic, many of them have been targets of “AI bombing”, the practice of using generative AI to rapidly create and post large volumes of fake, misleading, or low-quality reviews about a business, artificially manipulating its online reputation.

A competitor, a bad actor, anyone with a prompt, can now bury a business under thousands of reviews it never earned, and no model, human or otherwise, can reliably tell the fake from the real. When reputation can be manufactured at scale, provenance stops being a nice-to-have. It becomes the whole business.”

A simple test you can use

What I started with Epinions became an industry all its own, one that’s still serving, and potentially costing, the consumers who rely on it today. Before you act on a recommendation you find online, ask who’s telling you this and what they have to gain or lose. That question hasn’t changed since 1999. What’s changed is how much the industry profits when you don’t ask it.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.