Blog · 2026-07-20 · 8 min read
How an answer engine picks a winner — and the five ways you lose
When ChatGPT names three brands and stops, that verdict has an anatomy. Understand it and “why not us?” stops being a mystery and becomes a work item.
Ask an answer engine a buying question and watch what it actually does. With web search on, it issues queries, retrieves a handful of pages, reads them, and composes an answer — typically a few named options with reasons, drawn from what the retrieved pages said. The output feels like judgment. Mechanically, it's closer to synthesis: the engine amplifies the sources it retrieved and trusted in that moment.
That mechanism is good news, because everything in it is inspectable. Which pages got retrieved. Which brands those pages named. Whether your content was liftable into an answer. When you lose a question, the cause lives somewhere specific in that chain — and in our audits, losses sort cleanly into five gap types.
The five gaps
Content gap — the page that answers the question doesn't exist on your site. Engines can't cite what isn't written. This is the most common gap and the most fixable: publish the page, titled the way buyers actually phrase the question, saying the thing the answer needs to contain.
Format gap — the content exists but isn't liftable. A thousand words of narrative prose is hard for an engine to quote; a clean comparison table or a ranked list with criteria is easy. Same information, different machine-readability. Restructure and the engine can finally use what you already wrote.
Authority gap — the engine answered from a third-party “best of” roundup, and you're not in it. This is the gap brands feel most acutely, because the deciding page isn't theirs. The fix is narrow and surgical: the diagnosis names the exact roundup the engine cited, and that page — not a hundred hypothetical placements — is the target.
Freshness gap — the competitor's page is newer or visibly maintained, and grounded retrieval favors it. Updating content and its published date is unglamorous work that moves answers more than it has any right to.
Schema / entity gap — the engine can't confidently tell what you are. No structured data, ambiguous naming, no clear product entity. If the model can't classify you, it can't recommend you, no matter how good the prose is. Schema markup and a crisp entity definition are table stakes for being machine-legible.
Diagnosis has to carry evidence
A gap label without proof is just an opinion with a taxonomy. The diagnosis only becomes actionable when it arrives with the receipt: the stored answer the engine gave, the page it actually cited, and the classification of why that page beat yours. That's how “we lost the startup-expense question” turns into “get listed in this specific roundup” — a task someone can complete this sprint.
It also keeps everyone honest about what's promisable. Nobody controls a probabilistic engine, and anyone guaranteeing you a mention is guessing on your invoice. What's real: fix the named gap, re-run the same questions next month, and watch whether they flip from lost to won. Evidence in, work out, movement measured. That's the whole game — and right now, while most of your competitors haven't even looked, it's a game with remarkably little competition.
— The SolvedAgain team