Reputation Is the Signal AI Agents Trust

Two buyers approach the same kind of decision. One is a procurement manager choosing analytics software; the other is a shopper hunting for a winter coat. For a decade they behaved nothing alike. Now both have delegated the decision to an agent, and the agent does the same thing in either case: it goes looking for a reputation it can verify before it recommends anything at all.

The signal that survives that handoff is reputation. Ratings, reviews, the aggregated record of what other buyers actually experienced. When a model assembles a recommendation it leans hardest on the evidence it can independently trust, and a brand's own marketing is the evidence it discounts first.

Few people have watched this shift longer than Godard Abel, who co-founded G2 and built the largest software-review marketplace on a single conviction. He states it plainly in his 2026 year-in-review: "the buyer's voice is the ultimate truth." For fifteen years that meant verified peer reviews on a website people visited. In 2026 it means something sharper, because that voice has now become the data a model consults on the shopper's behalf.

Abel describes the discovery change in four words. The playbook "shifted this year from 'ranking' to 'answering.'" Rather than competing for a position on a results page, a brand now competes to be the answer a model returns. G2's reply was to launch a dedicated answer-engine-optimization category that grew from seven products to more than 150 in a single year, and to become, by its own measurement, the most cited B2B software source across large language models. A model cites G2 for the reason the marketplace always mattered: the reviews are verified, and verification is precisely what a reasoning system can rest a recommendation on.

The buyer changed jobs, and the trust test got stricter

A procurement agent and a shopping agent are engineered for different catalogs, yet they interrogate a product the same way. Can I trust what this listing claims? A human answered that with a blend of brand affinity, a salesperson's reassurance, and a gut read. An agent carries none of those instruments. It has the catalog in front of it and whatever external reputation it can locate and verify. That agent is a distinct buyer now, and it reasons toward a recommendation instead of absorbing a pitch.

In retail, the verified signal is the star rating

What Abel observed in software, retail research arrived at independently. Across more than 16,000 simulated shopping rounds spanning four AI models, star ratings were the one signal that reliably lifted an agent's selection, while most persuasion tactics produced erratic or negative effects. A rating is the consumer-facing version of the same principle: an aggregated, verified record of experience that a model can weigh without trusting the seller's framing. It carries a second advantage that matters enormously here. A rating already arrives in a structured, machine-readable shape, which is more than most of a catalog can claim.

The most trusted signal is the one you author least

The uncomfortable part for a brand is structural, and it runs against the grain of how most marketing teams have been trained to operate. The evidence a model trusts most is the evidence the brand controls least. You compose your own product description. You do not compose your own reviews, and a reasoning model registers that difference perfectly and weights the two sources accordingly. That hierarchy does not make authored content pointless. It makes authored content the other half of the assignment. A feed carries availability and little of the merchant's judgment, and how a brand describes its product has become its shelf placement. Reputation earns the agent's trust. Authored reasoning gives the agent something to reason with. A brand competing for the recommendation needs both working together.

What a software marketplace learned, a catalog can apply

Abel's company spent fifteen years proving that buyers trust other buyers more than they trust sellers. AI did not overturn that instinct. It hardwired it, because a model with no loyalty and no gut weights verified reputation above anything a brand asserts about itself. A retailer who files ratings under vanity metrics is overlooking the one signal the new buyer consults first. A retailer who treats reviews as core inventory, surfaced and structured for the agent beside fit, policy, and substitution logic, is handing the reasoning system exactly the evidence it trusts. Software or a sweater, the agent reasons the same way. It believes the buyers before it believes the brand.

Sumit Jagdale is the founder of Sartorial.