Skip to main content
Colter
Check my product

Check a public product page for free, without an account.

Recommendation Audit

Why Don’t My Shopify Products Show Up in Perplexity Recommendations?

A readiness score cannot tell you whether an answer engine recommends your product. Here is a bounded way to measure the miss, fix one evidence gap, and rerun the same buyer question.

Colter Team·

A technically complete Shopify product page can still be absent from an AI shopping answer.

We tested 15 public Shopify products on August 13, 2026. Each one got a single prewritten, unbranded buyer question in a logged-out Perplexity session. Thirteen products were absent from their rendered answers, and none of the 15 merchant domains was cited.

This is a small selected cohort with one observation per prompt, so it is not a market-wide miss rate. It does show where a page score stops answering the question.

What did the test actually measure?

Each question described a buyer job without naming the brand. For example:

What olive oil should I use to finish roasted vegetables?

An anonymized finishing-oil product page passed every deterministic product check in our audit. The Perplexity answer did not mention the target product and did not cite its merchant domain. It told the shopper to pick a fresh extra-virgin oil by flavor strength.

The result contains two separate facts:

  • MEASURED: the product page published the evidence our deterministic checks look for.
  • OBSERVED: this Perplexity answer did not mention or cite the product.

Nothing here explains the omission. The engine may have leaned on third-party sources, comparison coverage, freshness, geography, its own retrieval behavior, or evidence we never inspected.

Why didn’t a technically complete page earn a recommendation?

Structured product data makes the facts on the page machine-readable. A recommendation question asks for something the facts alone do not settle: which product fits this job, and how it compares to the obvious alternatives.

For the olive-oil prompt, a useful comparison would explain when this finishing oil is the right choice, how it differs from an oil meant for cooking, and what a shopper gives up by picking something else. Some of that evidence may also need to exist off the merchant site.

The same pattern held across the run. Eleven products that passed every deterministic check were still missing from their buyer answer. Two products were mentioned, and neither merchant domain was cited.

How should I audit a missing Shopify product?

One product, one buyer question, one named engine, all three fixed for the whole test.

  1. Record what the page publishes. Product identity, price, availability, identifiers, policies, and the facts the buyer question turns on.
  2. Record the answer. The exact prompt, engine, timestamp, products mentioned, merchant citations, and a hash of the answer text.
  3. Diagnose one gap. Separate facts missing from the product page from the comparison or authority evidence the answer appears to lean on.
  4. Make one bounded change. Correct or add truthful evidence. Change several things at once and the rerun will not tell you which one mattered.
  5. Rerun the same question. Record whether the product gained or lost a mention, gained a citation, moved in the comparison set, or stayed absent.

A better page score tells you the page improved. The rerun is the only part of this that speaks to the answer, and one rerun is still one observation.

What would count as a successful fix?

Success depends on the buyer job. For this anonymized example: an accurate mention of the product for finishing roasted vegetables, a correct distinction between finishing and cooking oil, or a citation to the merchant domain somewhere relevant in the answer.

Answer engines change their sources and their wording between runs. Watch the same small prompt set over time and keep each engine’s record separate.

FAQ

Does Product JSON-LD make a Shopify product appear in Perplexity?

It makes the facts easier to read off the page. It does not settle whether an engine mentions, recommends, ranks, or cites the product.

Start with questions tied to a real product and a real buyer job. A useful page answers the question directly, publishes accurate machine-readable facts, and shows where its claims come from. Volume without proof is more pages to maintain.

Can the same audit be used for ChatGPT or Google AI Mode?

Yes, with a separate observation for each engine. Do not blend engines into one “AI visibility” score.

What does Colter Recommendation Audit do?

It reads one public Shopify product page and tells you what to improve first. You can optionally save a buyer question and compare it later against a real answer you capture yourself. The saved question does not change the checked-page average or which fix comes first.

Run a Recommendation Audit

Test scope and evidence

  • Date: August 13, 2026
  • Engine: Perplexity Search, logged out, public default experience
  • Cohort: 15 selected public Shopify products
  • Method: one prewritten unbranded category prompt per product; one observation per prompt
  • Observed: 13 products omitted; two mentioned; zero target merchant-domain citations
  • Product state: 13 pages passed deterministic product checks; 11 of those were absent
  • Fix boundary: one actionable page-fix candidate identified; no merchant change or unchanged-prompt rerun performed
  • Limitation: this does not establish stable placement, cross-engine behavior, causation, traffic, conversion, or willingness to pay

The underlying answer records are retained internally. Public observations are anonymized to avoid identifying the merchants or products.

← All posts