Signals Database
Signals Database
The structured catalog of the signals that shape which store, product, or offer an AI shopping agent selects. Every record is evidence-typed (spec-fact, reported, or hypothesis) with a confidence level and, where a platform documents its behavior, honest per-platform status.
The whole database is also published in machine-legible form: /signals.json (structured) and /signals.md (Markdown catalog).
Showing all 39 signals.
| Signal | Affects | Platforms documented | Evidence | Confidence | Impact | Last verified |
|---|---|---|---|---|---|---|
| Agent infrastructure reachability | Discovery | 3/5 ChatGPT, Gemini, Claude | Hypothesis | Medium | 3/3 | |
| Content equivalence (no cloaking) | Discovery | 2/5 Gemini, Copilot | Spec-fact | Very high | 3/3 | |
| Crawl directives (robots.txt + sitemap) | Discovery | 4/5 ChatGPT, Gemini, Perplexity, Claude | Spec-fact | Very high | 3/3 | |
| Market, language and currency eligibility | Discovery | 3/5 ChatGPT, Gemini, Copilot | Spec-fact | Very high | 3/3 | |
| Agent-readiness monitoring and ownership | Discovery | 1/5 Gemini | Hypothesis | Low | 2/3 | |
| llms.txt priority ordering | Discovery | None documented | Spec-fact | Very high | 2/3 | |
| Machine-readable page representation | Discovery | None documented | Hypothesis | Low | 2/3 | |
| Origin serve-ability for agents | Discovery | 3/5 ChatGPT, Gemini, Claude | Hypothesis | Low | 2/3 | |
| Track AI-answer presence, not just referrals | Discovery | 1/5 Gemini | Hypothesis | Low | 2/3 | |
| Conversational product attributes | Matching | 1/5 Gemini | Spec-fact | Very high | 3/3 | |
| Offer markup completeness and truthfulness | Matching | 2/5 ChatGPT, Gemini | Spec-fact | Very high | 3/3 | |
| Server-rendered, crawlable content | Matching | 2/5 Gemini, Claude | Hypothesis | Low | 3/3 | |
| Answering agent tool calls | Matching | 3/5 ChatGPT, Gemini, Claude | Spec-fact | High | 2/3 | |
| Product identity and canonical URL | Matching | 3/5 ChatGPT, Gemini, Copilot | Spec-fact | Very high | 2/3 | |
| Validation hygiene (use existing validators) | Matching | 2/5 Gemini, Copilot | Spec-fact | Very high | 2/3 | |
| Variants and titles use the words shoppers type | Matching | 3/5 ChatGPT, Gemini, Claude | Hypothesis | Low | 2/3 | |
| Competitive price awareness with parity | Comparison | 2/5 ChatGPT, Gemini | Reported | High | 3/3 | |
| Delivery ETA is computable per destination | Comparison | 4/5 ChatGPT, Gemini, Copilot, Claude | Hypothesis | Low | 3/3 | |
| Descriptions answer relayed buyer questions | Comparison | 2/5 ChatGPT, Gemini | Hypothesis | Low | 3/3 | |
| Facts per token, complete against peers | Comparison | 2/5 ChatGPT, Gemini | Hypothesis | Low | 3/3 | |
| Offer cost legibility | Comparison | 2/5 ChatGPT, Gemini | Hypothesis | Low | 3/3 | |
| Loyalty benefits visible to the agent | Comparison | 1/5 Gemini | Spec-fact | Very high | 2/3 | |
| Query-conditional rewrites are a targeted experiment | Comparison | 1/5 ChatGPT | Reported | High | 2/3 | |
| Return terms are fully legible | Comparison | 4/5 ChatGPT, Gemini, Copilot, Claude | Hypothesis | Low | 2/3 | |
| ACP feed and checkout readiness | Choice | 1/5 ChatGPT | Spec-fact | Very high | 3/3 | |
| Agent-payment rail readiness | Choice | 3/5 ChatGPT, Gemini, Claude | Reported | High | 3/3 | |
| Audit position and ordering effects per engine | Choice | 1/5 ChatGPT | Reported | Medium | 3/3 | |
| Brand familiarity versus a small rating edge | Choice | None documented | Reported | Medium | 3/3 | |
| Defined agent win-rate measurement | Choice | None documented | Hypothesis | Low | 3/3 | |
| Earn endorsement, never self-label sponsored | Choice | None documented | Reported | High | 3/3 | |
| Maker or primary seller status | Choice | 1/5 ChatGPT | Spec-fact | Very high | 3/3 | |
| Offer consistency and freshness across surfaces | Choice | 3/5 ChatGPT, Gemini, Copilot | Hypothesis | Low | 3/3 | |
| Re-test after every major model update | Choice | 1/5 ChatGPT | Reported | High | 3/3 | |
| Review depth: count, recency, distribution | Choice | 2/5 ChatGPT, Gemini | Hypothesis | Low | 3/3 | |
| Guest checkout completability | Choice | 1/5 ChatGPT | Hypothesis | Low | 2/3 | |
| Personalization context you cannot see | Choice | 1/5 ChatGPT | Spec-fact | Very high | 2/3 | |
| Price sensitivity varies by model | Choice | None documented | Reported | High | 2/3 | |
| Seller identity reachable in two hops | Choice | 2/5 ChatGPT, Copilot | Hypothesis | Low | 2/3 | |
| UCP profile and order-management readiness | Choice | 2/5 Gemini, Copilot | Spec-fact | Very high | 2/3 |
Platform coverage matrix
Signals as rows, shopping surfaces as columns. A cell is filled only where a platform documents its own behavior (69 of 195); everything else readsunknown, which is the honest default and not a gap waiting to be filled in.
The shape of what is filled says more than the count. Google and OpenAI document feed and structured-data requirements in detail, so the signals about being eligible are well covered. Almost nothing is documented about how any engine ranks one eligible product against another. Microsoft documents a feed, Anthropic publishes a reference implementation that contains no ranking logic, and Perplexity documents the product data its Merchant Program expects and little else. The blank half of this table is the subject of this site.
This view is generated from the signal records; fixing a cell means fixing a signal, never this table.
| Signal | ChatGPT | Gemini | Perplexity | Copilot | Claude |
|---|---|---|---|---|---|
| Agent infrastructure reachability | Partial | Partial | Unknown | Unknown | Partial |
| Content equivalence (no cloaking) | Unknown | Supported | Unknown | Reported | Unknown |
| Crawl directives (robots.txt + sitemap) | Supported | Supported | Supported | Unknown | Supported |
| Market, language and currency eligibility | Supported | Partial | Unknown | Supported | Unknown |
| Agent-readiness monitoring and ownership | Unknown | Partial | Unknown | Unknown | Unknown |
| llms.txt priority ordering | Unknown | Unknown | Unknown | Unknown | Unknown |
| Machine-readable page representation | Unknown | Unknown | Unknown | Unknown | Unknown |
| Origin serve-ability for agents | Partial | Partial | Unknown | Unknown | Partial |
| Track AI-answer presence, not just referrals | Unknown | Partial | Unknown | Unknown | Unknown |
| Conversational product attributes | Unknown | Supported | Unknown | Unknown | Unknown |
| Offer markup completeness and truthfulness | Partial | Supported | Unknown | Unknown | Unknown |
| Server-rendered, crawlable content | Unknown | Supported | Unknown | Unknown | Supported |
| Answering agent tool calls | Partial | Partial | Unknown | Unknown | Partial |
| Product identity and canonical URL | Supported | Supported | Unknown | Supported | Unknown |
| Validation hygiene (use existing validators) | Unknown | Supported | Unknown | Partial | Unknown |
| Variants and titles use the words shoppers type | Partial | Supported | Unknown | Unknown | Partial |
| Competitive price awareness with parity | Supported | Partial | Unknown | Unknown | Unknown |
| Delivery ETA is computable per destination | Partial | Supported | Unknown | Partial | Partial |
| Descriptions answer relayed buyer questions | Partial | Partial | Unknown | Unknown | Unknown |
| Facts per token, complete against peers | Partial | Partial | Unknown | Unknown | Unknown |
| Offer cost legibility | Partial | Partial | Unknown | Unknown | Unknown |
| Loyalty benefits visible to the agent | Unknown | Supported | Unknown | Unknown | Unknown |
| Query-conditional rewrites are a targeted experiment | Partial | Unknown | Unknown | Unknown | Unknown |
| Return terms are fully legible | Partial | Supported | Unknown | Partial | Partial |
| ACP feed and checkout readiness | Supported | Unknown | Unknown | Unknown | Unknown |
| Agent-payment rail readiness | Partial | Partial | Unknown | Unknown | Unsupported |
| Audit position and ordering effects per engine | Reported | Unknown | Unknown | Unknown | Unknown |
| Brand familiarity versus a small rating edge | Unknown | Unknown | Unknown | Unknown | Unknown |
| Defined agent win-rate measurement | Unknown | Unknown | Unknown | Unknown | Unknown |
| Earn endorsement, never self-label sponsored | Unknown | Unknown | Unknown | Unknown | Unknown |
| Maker or primary seller status | Supported | Unknown | Unknown | Unknown | Unknown |
| Offer consistency and freshness across surfaces | Partial | Supported | Unknown | Partial | Unknown |
| Re-test after every major model update | Partial | Unknown | Unknown | Unknown | Unknown |
| Review depth: count, recency, distribution | Partial | Supported | Unknown | Unknown | Unknown |
| Guest checkout completability | Supported | Unknown | Unknown | Unknown | Unknown |
| Personalization context you cannot see | Supported | Unknown | Unknown | Unknown | Unknown |
| Price sensitivity varies by model | Unknown | Unknown | Unknown | Unknown | Unknown |
| Seller identity reachable in two hops | Partial | Unknown | Unknown | Partial | Unknown |
| UCP profile and order-management readiness | Unknown | Supported | Unknown | Supported | Unknown |