Complete sample report

A complete report, exactly as delivered

This is the complete, unabridged report for a real store in our benchmark: a specialty aquarium-decor store, audited on 2026-08-17. Only the store's name and domain are masked; every number, finding, and fix is exactly what the audit produced. A paid report follows exactly this shape for your own store.

Composite score

75.6/100

Standing: top 8.4% of 149 stores measured

This run's AI-answer providers: anthropic STUB; gemini LIVE; openai ERROR; perplexity STUB; serper STUB. LIVE means a real provider call this run; STUB means that provider was not queried; ERROR means our call to it failed on our side.

Dimension detail

AI Answer Presence: 100.0/100

AI Answer Accuracy (own-domain citation): 77.8/100

Structured Data Completeness: 50.0/100

Discoverability Hygiene: 100.0/100

Reputation Signals: 50.0/100

Every finding this run surfaced

HIGH AI answers about the brand did not cite the brand's own domain
A shopper who asks that question is ready to buy. When the assistant's answer skips your store, the visit goes to whichever competitor it names, and no analytics report will ever show you the loss.
Measured: gemini's answer to "best products online store" did not cite the store's own domain (2 of 9 live probes affected).
The fix: Improve on-site authority for brand/product questions (canonical URLs, no paywall or JS-only rendering of key content, clear brand name usage in page titles and headers) so AI engines cite your own site instead of a third-party page when answering about you.
MEDIUM Partial schema.org markup
Product markup is how an assistant reads your catalog with certainty. Without it your products get described secondhand, from whatever a crawler could scrape.
Measured: Only Organization JSON-LD found; Product/ProductGroup markup missing.
The fix: You have one of Product/ProductGroup or Organization JSON-LD but not both -- add the missing type so AI crawlers get a complete picture of your storefront in one pass instead of a partial one.
MEDIUM No Wikidata entity or Wikipedia page found
Answer engines check primary sources to decide whether a brand is real before they recommend it. A store with no entry starts every trust question one citation behind a competitor that has one.
Measured: No Wikidata entity or Wikipedia page found for this brand.
The fix: Create (or claim) a Wikidata entity for your brand and, where notability supports it, a Wikipedia page -- these are primary sources AI answer engines cite for "who is this company" questions.
LOW Not enough surfaces to verify NAP consistency
Consistent public contact details are one of the cheapest trust signals an answer engine can verify. When it cannot verify them, that signal is absent from every answer about you.
Measured: Fewer than 2 surfaces carried an extractable phone number (found on: none); cannot verify consistency.
The fix: Publish a phone number on at least two public pages (e.g. homepage plus a dedicated contact page) so NAP consistency can be verified by crawlers and by future audits.

Other categories probed this run

llms.txt: already clean (check passed).
robots.txt / sitemap: already clean (check passed).
Review signals: already clean (check passed).
AI answer presence: already clean -- every live probe returned an answer.

Methodology

The AI-answer layer of this audit is measured via official provider APIs -- OpenAI (Responses API `web_search` tool), Anthropic (Messages API `web_search_20250305` server tool), Perplexity (`sonar` chat- completions API), Serper.dev (a SERP API used as a proxy for Google AI Overviews), and Gemini (Generative Language API with the `google_search` grounding tool) -- NOT consumer free-tier chat sessions. These five named providers are queried with the same battery of brand/product/trust questions a prospective customer might ask an AI assistant.

Because these are API-layer calls rather than a capture of any specific consumer chat product's live output, the results are a PROXY for the consumer-facing AI-search answer layer rather than a direct, guaranteed match to what any individual person sees in ChatGPT, Claude, or a Google search on a given day. Answer-layer results can vary by session, model version, and provider-side ranking changes outside this tool's control.

Each provider is key-gated: when no API key is configured for a provider, that provider runs in STUB mode (a canned response, zero network, zero cost) instead of LIVE mode (a real network call). Every scored audit records which mode each of the five providers ran in for that specific run, so a report reader can tell a canned-stub result from a real live probe at a glance -- a report is never presented as a full live audit when one or more providers ran in stub mode.

The deterministic web-presence checks (schema.org markup, llms.txt, robots.txt/sitemap health, Wikidata/Wikipedia footprint, NAP consistency, review signals) are plain, unauthenticated HTTP requests against the target's own public surfaces and carry no proxy caveat -- they measure exactly what they claim to measure.

This is what a report looks like. See pricing for your store