Why two businesses with the same facts get different recommendations.

Confidence is contagious across a page, in both directions

Say two shops both sell a similar waterproof jacket, both stating the same 10,000mm hydrostatic rating, both in stock, both at a similar price. An AI agent asked to recommend one for a wet-weather walking trip still has to pick, and it doesn't flip a coin. It looks at which shop's claim is easier to verify quickly, which one has a source attached rather than just a number on a page, and which one has nothing else on the page contradicting or hedging the claim. Identical facts, unequal confidence.

This is easy to miss from inside a business, because the facts genuinely are the same, and it feels like the recommendation should be too. What varies is everything around the fact: whether the rating is stated once, clearly, or repeated inconsistently in three places with slightly different numbers; whether the product page also carries vague, unverifiable claims nearby that make the whole page read as less careful; whether the stock and size information is current or plausible-looking but unconfirmed.

An AI weighing one clear, sourced claim against a page full of vague ones doesn't treat the clear claim as an island. A shop that's precise about the hydrostatic rating but vague about sizing, vague about returns, and vague about stock reads as inconsistently careful, which drags down confidence in the one claim that was done properly. The reverse is also true: a shop that's specific and current everywhere earns a kind of blanket trust that makes each individual claim easier to act on.

What breaks the tie

Three things, in practice: consistency, so the same fact reads the same wherever it appears rather than drifting slightly between the product page, the size guide, and the FAQ; freshness, so a stock or price claim reflects today rather than whenever the page was last edited; and completeness, so the facts an AI needs to complete a recommendation, not just describe the product, are present rather than assumed. A shop that's strong on all three wins the tie even against an identical fact stated somewhere less carefully.

Say the product page states the hydrostatic rating clearly, but the size guide, updated separately eighteen months ago, still describes an older cut that ran differently. Neither page is wrong on its own terms. Together they're a shop whose facts don't quite agree with themselves, which an AI checking both pages has no way to resolve except by trusting the whole shop a little less than one whose two pages say the same thing.

Why this is worth fixing before assuming the product itself needs to change

It's tempting to read an uneven recommendation as a signal the product isn't good enough, when the actual gap is often just how carefully, consistently, and currently the facts around it are kept. That distinction matters, because fixing consistency and freshness is genuinely achievable work, while assuming the product needs to change is a much bigger, often unnecessary response to the same problem. That's the layer Selfe exists to build: keeping a shop's real facts consistent and current everywhere an AI might check them, so two identical products get compared on their actual merits, not on which shop happened to keep tidier records.

How would we notice this is happening to us?

Ask an AI to compare your product against a close competitor's on a shared spec. If it hedges on yours and not theirs, the gap is usually in how the fact is presented, not the fact itself.

Does this mean we need identical formatting to a competitor?

No, just internal consistency: the same fact stated the same way everywhere it appears on your own site, with a current date behind anything that changes.