Why Doesn’t AI Recommend My Store? A Diagnostic Guide
AI may omit your store for several unrelated reasons: the tested surface could not access or select the page, the prompt intent did not fit, commerce facts conflicted, the entity was ambiguous, evidence was weak, or the answer simply varied. Diagnose the exact surface and observation before changing the site.
Do not begin by asking which schema or FAQ to add. Begin with the observed failure: surface, exact prompt, date, locale, account state, answer, citations, and accessible pages. That evidence narrows the problem before theme edits, content production, or vendor dashboards create more variables.
The goal is not to force a recommendation. It is to remove observable barriers, align public facts with buyer intent, and build a repeatable record of what each system returned. Selection remains conditional on systems and contexts the store does not control.
Why isn’t “AI does not recommend us” one bug?
“AI does not recommend us” is a symptom, not a diagnosis. A missing answer can originate in access, discovery or index eligibility, intent mismatch, incomplete or conflicting commerce facts, ambiguous entities, limited independent evidence, or ordinary answer volatility. Fixing one eligibility layer may remove a barrier, but it cannot guarantee retrieval, citation, recommendation, ranking, or sales.
| Failure layer | Observable symptom | Verify next |
|---|---|---|
| Surface | One engine or mode misses | Exact conditions |
| Access | Blocked, challenged, or empty | Logs, status, HTML |
| Eligibility | Noindex or canonical conflict | Controls and discovery |
| Intent | Branded found; buyer prompt absent | Prompt-page fit |
| Commerce facts | Price, stock, policy conflict | Visible facts and markup |
| Entity | Names or ownership unclear | Organization consistency |
| Evidence | Claims lack support | Sources, reviews, policies |
| Volatility | Answer changes | Repeated fixed panel |
Which answer surface are you actually testing?
First name the surface: ChatGPT Search, Perplexity, Google AI features, and shopping or product experiences are not interchangeable tests. They use different retrieval systems, crawler controls, indexes, interfaces, personalization, and product data. Record the engine, mode, prompt, locale, account state, date, answer, citations, and source URLs before comparing results.
Read OpenAI’s ChatGPT Search explanation, Perplexity’s bot controls, and Google’s AI-features guidance as surface-specific documentation. Shopping tests also need the exact catalog, offer state, and product surface. See how to show up on ChatGPT for a bounded workflow.
Can the relevant crawler retrieve the page?
Verify access with the crawler that matters to the surface. OpenAI distinguishes OAI-SearchBot, used for search inclusion, from GPTBot, used for training controls. A permissive robots rule does not override a CDN, WAF, login wall, rate limit, or JavaScript failure. Confirm a 200 response in logs and useful visible HTML.
Use OpenAI’s bot documentation, Google’s robots.txt introduction, Shopify’s robots.txt guide, and the AI crawler checker.
Crawl checklist:
- Confirm the relevant user agent and robots permission.
- Remove CDN, WAF, rate-limit, and login challenges.
- Follow redirects to the final 200 response.
- Compare initial and rendered useful HTML.
- Find user-agent-specific 200 events in server logs.
Are discovery and index eligibility aligned?
Google says pages appearing in its AI features follow normal Search eligibility and foundational requirements; no special AI schema is required. Check indexability, canonical alignment, internal links, sitemaps, and the page’s ability to satisfy Search Essentials. Submission or crawler permission can aid discovery, but neither compels indexing, an AI appearance, or a recommendation.
Use Google’s AI-features guidance with Search Essentials. Inspect Google-specific eligibility in those sources rather than treating a third-party visibility score, crawler permission, or sitemap submission as proof of selection.
Are product facts, schema, and entities consistent?
Accurate Product, Offer, Organization, and Breadcrumb markup can clarify visible commerce facts, but it cannot guarantee citation or recommendation. Match identifiers, variants, price, currency, availability, seller, brand, shipping, returns, and policies to the rendered page. Never invent Person expertise, reviews, AggregateRating, stock, test results, observed competitors, or unsupported product claims.
Compare Google’s Product guidance and Organization guidance with Schema.org Product and Offer. Shopify’s product object can expose theme data, but rendered facts still need verification. StoreCited remains the Organization entity.
Does content match non-branded buyer intent and show evidence?
Branded verification prompts and non-branded buyer prompts answer different questions. A branded prompt checks whether systems can recognize facts about your store; a non-branded prompt tests whether your page fits a buyer need among alternatives. Build content around genuine tasks, constraints, comparisons, proof, and policies, then review independent evidence without manufacturing mentions or reviews.
Google’s helpful content guidance supports useful, people-first material rather than an arbitrary word count.
Content and evidence checklist:
- State what the product is and who it fits.
- Keep variants, specifications, compatibility, and materials current.
- Show price, stock, shipping, returns, and restrictions.
- Explain comparison criteria, tradeoffs, and limitations.
- Attach sources, dates, and methods to consequential claims.
- Use only genuine, visible, independently traceable reviews.
What is the fastest triage sequence?
Use a fixed triage order so later tests rest on earlier evidence. Start by defining the failed surface and observation, then prove crawler access, resolve discovery and index contradictions, align commercial facts and entities, repair intent coverage and evidence, and finally repeat the same prompt panel. Do not jump from one missing answer to broad content production.
- Define the surface, prompt, mode, locale, account, date, and observed answer.
- Verify the relevant crawler’s robots permission, final status, logs, and visible HTML.
- Resolve noindex, canonical, redirect, internal-link, and sitemap contradictions.
- Align visible product facts, policies, markup, and Organization identity.
- Improve buyer-intent coverage, claim support, comparisons, and limitations.
- Repeat the fixed branded and non-branded panel; label every change.
How should changes be measured?
Measure answer visibility with a fixed, repeated panel rather than isolated screenshots. Preserve exact prompts, engines, modes, locales, account states, dates, answers, citations, and source URLs. Track branded and non-branded sets separately. Compare like with like over time, and keep Google Search data, answer observations, and business outcomes separate because correlation is not causation.
Use the AI visibility measurement guide to document a reproducible panel.
| Evidence stream | Record | Cadence | Cannot prove |
|---|---|---|---|
| Access | User agent, status, log, HTML | After changes | Selection |
| Google Search | Clicks, impressions, position | Weekly or monthly | AI citation |
| Answer surfaces | Exact outputs, citations, URLs | Fixed schedule | Complete coverage |
| Business | Sessions, leads, orders | Reporting period | Cause |
Where does StoreCited fit?
StoreCited fits only at the readiness stage. It performs a point-in-time scan of observable public Shopify and DTC storefront signals and returns prioritized gaps. It does not monitor live answers, expose proprietary model indexes, generate publish-ready schema or FAQs, identify models’ actual competitors, change eligibility, or guarantee selection. Treat it as a dated diagnostic.
StoreCited can help decide what to inspect next, but it does not replace platform-specific observation. Run the free StoreCited readiness scan, repair supported gaps, then verify access and outcomes in the systems that produced the original evidence.
Get the answer for your specific store