Preferd← Back to home
Strategy

Why AI assistants recommend your competitor (and not you)

Here's a two-minute experiment that has ruined a few merchants' mornings: open ChatGPT or Perplexity and ask the question your customers ask — “best [your product category] for [your typical use case]?”. If a competitor's product comes back and yours doesn't, you've just watched a sale you never knew you were losing. The good news: when we dig into why a model picked the rival, the reasons are boringly consistent — and fixable.

Reason 1 — Their data answers the question; yours doesn't

The model matched “keeps drinks hot on winter hikes” to a product whose title and description literally say “24 h heat retention” and “designed for sub-zero conditions”. Your equivalent product may be objectively better — but if the description is two lines and the title is “Alpine Bottle — NEW”, the model has nothing to match and nothing to quote. Assistants recommend what they can defend with specifics.

Reason 2 — Your product isn't a verifiable entity

Models favor products they can corroborate: a real brand name, a GTIN that links to reviews elsewhere, a taxonomy category that confirms what the thing is. A catalog with “Default” as vendor, no barcodes and uncategorized products reads — to a machine — like a store that might not ship. The rival with boring-but-complete data wins the trust tiebreak every time.

Reason 3 — The web agrees on them, not on you

Search-grounded assistants blend your product data with what the wider web says. A competitor mentioned in two buying guides and thirty reviews has third-party corroboration you can't fake with on-site edits. This one is the slowest to fix — but note that it compounds: being recommended by assistants drives traffic that produces the reviews that drive more recommendations.

Reason 4 — You're both in the feed; they win the comparison

Since Shopify syndicates eligible catalogs to AI surfaces by default, “being in AI” stopped being a differentiator — everyone eligible is in. The model isn't choosing between present and absent stores; it's ranking within a crowded shelf. Feed-visible attributes — category metafields, images, identifiers — decide those rankings, and they're exactly the fields most catalogs leave empty.

Reason 5 — Nobody on your side is measuring

Your competitor may simply have noticed first. Merchants who track category answers month over month catch regressions (“we dropped out of the espresso-grinder answers in June”) and fix them; merchants who don't measure find out from their revenue, quarters later. Measurement is the cheapest advantage on this list.

None of these five reasons is “they installed a magic file”. Technical plumbing (llms.txt, schema) is table stakes — the recommendation gap is a data-quality and evidence gap.

The fix, in order