Glossary
16 terms for measuring AI crawlers, AI referral traffic and AI visibility, each defined in a sentence or two. Every entry shows how the term appears in Traceten and links to the product page that covers it, from answer fetches and crawler verification to detection confidence, share of voice and query fan-out.
An AI crawler is an automated client that an AI company runs to fetch web pages, whether to train a model, build a search index or answer a question in real time. It names itself in its user agent and runs no JavaScript, so a browser analytics tag never sees it.
In Traceten: Traceten's directory lists 60 crawlers from 21 providers. Their hits reach Traceten from your server or Cloudflare, not from the page.
An answer fetch is a request an AI assistant makes in real time to read a page while it answers a question someone just asked. ChatGPT-User, Claude-User and Perplexity-User are answer-fetching crawlers.
In Traceten: Traceten files these under Answers, one of four crawler categories. It is the closest observable sign that your page was read for a live answer.
A search index crawler fetches pages continuously to build and refresh the index an AI search product answers from. OAI-SearchBot, Claude-SearchBot and PerplexityBot are search index crawlers.
In Traceten: Traceten files these under Search index, and its robots.txt checker reports each one apart from the same vendor's training crawler, such as OAI-SearchBot beside GPTBot.
A training crawler collects pages into a corpus used to train AI models. GPTBot, ClaudeBot and CCBot are training crawlers.
In Traceten: Traceten counts these under Training, so they never mix with the answer fetches that read a page for a live question.
A robots.txt token is the name a crawler answers to on a robots.txt User-agent line. Most tokens match a crawler's user agent, but a few, such as Google-Extended and Applebot-Extended, are permission tokens that control how content is used and never send a request.
In Traceten: Because a permission token never fetches anything, Traceten marks a live request that carries one as spoofed.
A spoofed crawler is a request that claims a crawler's user agent but comes from outside the network its vendor publishes. Anyone can type GPTBot into a user agent, so the claim alone proves nothing.
In Traceten: Traceten marks a hit spoofed when the claimed vendor publishes its IP ranges and the address falls outside them, or when it carries a permission token that never fetches, and keeps those hits apart from genuine ones.
Crawler verification is checking that a request claiming to be a crawler really comes from its vendor. The strongest checks match the address against the vendor's published IP ranges or confirm it with reverse DNS.
In Traceten: Every crawl in Traceten records its tier and a confidence: IP range 0.99, reverse DNS 0.95, ASN match 0.85, user agent only 0.60 and spoofed 0. User agent only is normal for a vendor that publishes no ranges.
An AI referral is a visit that arrives from an AI assistant, usually after someone clicks a link in an answer from ChatGPT, Claude, Perplexity or another assistant.
In Traceten: Traceten names the assistant from the referrer, UTM tags and user agent, across 25+ assistants, and records a confidence score and method on each visit.
The AI source names which assistant sent a visit, such as ChatGPT or Claude. The source kind names what category of traffic it was: ai, non_ai or unknown.
In Traceten: In Traceten they are separate axes, so a figure from one is never added to a figure from the other. Revenue adds unmatched as a fourth kind, for an order no visit could be tied to.
Detection confidence is a score from 0 to 1 that says how sure a classifier is about a visit's source. It is a degree of certainty, not a yes-or-no label.
In Traceten: A rule match on a known AI referrer, UTM tag or user agent records 0.99 in Traceten, and a visit nothing identifies records 0.
The detection method records which signal identified a visit's source, so every number can be traced back to its evidence.
In Traceten: Traceten stores the method on every visit, for example referrer_match, utm_match or user_agent_match on an AI visit, non_ai_referrer or non_ai_utm on other channels, and unknown when nothing matched.
A mention is an AI answer naming your brand. A citation is an AI answer linking to one of your pages. An engine can do either without the other.
In Traceten: In Traceten's AI visibility (Beta), presence counts mentions and citation rate counts answers that cite your domains. The two are never blended into one score.
Share of voice is your brand's mentions as a share of all mentions of the brands you track, across the AI answers sampled.
In Traceten: In Traceten's AI visibility (Beta) it is a share of mentions, not of answers: one answer that names three brands adds to all three.
Query fan-out is an answer engine splitting one question into several related searches and building its answer from pages found for any of them. It is why a page can be cited for a question it never targeted.
In Traceten: Traceten's cited-pages view (Beta) lists which of your pages the engines cited, joined to the answer fetches and AI visits each page received.
Generative engine optimization (GEO) is the practice of shaping content so AI-generated answers use and cite it. Its results show up in answers rather than in a ranked list of links.
In Traceten: Traceten measures it in three layers: the AI crawlers that read your pages, citations in AI answers (Beta), and the AI referral visits and revenue that follow.
Answer engine optimization (AEO) is writing pages so an answer engine can lift a passage and quote it: the answer first, sections that stand alone, and specific facts. It overlaps with GEO, with the weight on the quoted passage.
In Traceten: Traceten shows the outcome per page: which pages AI crawlers fetch, by purpose, and how many AI referral sessions each page gets.
No calls, no credit card.
