What happens to reviews once agents are the ones reading them.
A star rating tells a person something useful at a glance. It tells an AI agent almost nothing it can act on. As more of a property's reviews get read by agents rather than skimmed by people, what makes a review useful is shifting, and a 4.6 average isn't going to be the part that matters.
6 min read
What a star rating was always a proxy for
A small guesthouse has three hundred reviews and a 4.6 average, and has always treated that number as the headline. A person skimming reviews reads the average, glances at a couple of recent ones, and decides. An agent checking the same reviews on behalf of someone with a mobility need doesn't care about the average at all. It's looking for one specific, verifiable claim: whether previous guests confirmed step-free access to the room type being asked about. The average doesn't answer that. A handful of the actual reviews might.
A 4.6 average has always been a compression of a lot of specific facts into one number, comfortable beds, a helpful host, decent breakfast, a slightly awkward parking situation, averaged into something a person could skim in a second. That compression was useful for a person deciding quickly. It throws away almost everything an agent needs, because an agent isn't deciding quickly on a vague impression, it's checking whether one specific, stated requirement gets confirmed.
What becomes more valuable
Specific, checkable claims inside individual reviews start mattering more than the aggregate score sitting above them. A review that says the walk-in shower had a proper grab rail is doing real work for the next guest checking that exact thing. A review that just says "lovely stay, highly recommend" is doing almost none, for an agent or, if it's honest, for a person weighing a specific need either. The reviews that were always the most useful, the specific ones, are the ones an agent can use. The vague ones were always closer to noise.
The honest complication
None of this means a business can go collect specific-sounding reviews and expect an agent to treat them as settled fact. A review is one guest's account, not verified evidence, and treating it as more than that would be overclaiming what a review can prove. What a good review does is point toward something worth checking, not replace the checking. The distinction between a claim and a verified fact doesn't disappear just because the claim happens to be specific.
Where this shows up for a business
Selfe treats a review as one input among several, weighed against a business's own verified facts rather than standing in for them, so a specific claim in a review gets the credit it deserves without being mistaken for something a business itself confirmed.
Should we stop displaying our average star rating?
No, it still serves people skimming quickly. It just isn't what an agent is checking, so it shouldn't be the only thing a business relies on.
Can we just ask guests to write more detailed reviews?
It helps, genuinely, but a detailed review is still one account, not a verified fact, and shouldn't be treated as more certain than it is.
The next twelve months in agentic commerce.
Most writing on where AI agents are heading falls into one of two camps: everything changes overnight, or nothing really changes at all. Neither is honest. Some things about agentic commerce are already settled enough to build for. Others genuinely aren't decided yet, and pretending otherwise doesn't help anyone plan.
The research an agent does before you ever open a search box.
A growing share of product research now happens inside an ongoing conversation with an AI agent someone already trusts, rather than a fresh search each time. Nobody has a precise figure for how much yet, and any business that says otherwise is guessing. The direction, though, is real enough to plan around.
What changes for discovery in the next twelve months.
Not a sweeping prediction. Over the next year, the practical change for discovery is narrower and more useful than "everything will be different": more of the requests reaching a business will already carry real constraints attached, rather than arriving as a vague, browsable query.