What is T* (Topicality)?
T* (Topicality) is Google's base measure of how relevant a document is to the words in a query. The term comes from a February 18, 2025 call with Google engineer HJ Kim, recorded by the DOJ and filed as exhibit PXR0356 in the remedies phase of the Google search antitrust case.
It is a query dependent score. That separates it from quality scores such as Q*, which the same exhibit describes as largely independent of the query.
ABC signals are the key components of topicality (or a base score), which is Google's determination of how the document is relevant to the query.
HJ Kim, Google (as recorded by the DOJ), February 18, 2025 Call with Google Engineer HJ Kim (PXR0356), US Department of Justice
How does T* combine anchors, body text and clicks?
The exhibit names three inputs, which Search Engine Land summarised on May 13, 2025 as follows:
The exhibit says the three were developed by engineers and combined by hand rather than learned end to end by a model. The practical reading: relevance is triangulated. What the page says, what other pages say about it when they link, and how searchers behave after clicking all feed the same score.
- A, anchors: the text of links pointing at the document.
- B, body: the terms in the document itself.
- C, clicks: user interaction data, including how long a searcher stays on a result before returning to the results page.
T* (Topicality) effectively combines (at least) these three signals in a relatively hand-crafted way.
HJ Kim, Google (as recorded by the DOJ), February 18, 2025 Call with Google Engineer HJ Kim (PXR0356), US Department of Justice
What evidence is there that Google uses T*?
The evidence is trial evidence, not a Google announcement. Two exhibits from the remedies phase of US v. Google describe Google's ranking signals: the call with HJ Kim (PXR0356) and a January 31, 2025 call with Pandu Nayak (PXR0357). Both are DOJ records of interviews, so the wording is the DOJ's note of what was said, not a verbatim transcript or Google documentation.
Google has not published anything naming T*. Nothing in the exhibits gives weights, formulas or how the ABC signals are normalised, so any claim to a precise T* model is speculation.
Google uses over 100 raw signals.
Pandu Nayak, Google (as recorded by the DOJ), January 31, 2025 Call with Google Engineer Pandu Nayak (PXR0357), US Department of Justice
What does T* mean for my site?
Make the three inputs agree. If a page's body copy, the anchor text of internal links pointing to it and the queries it actually receives clicks for all describe the same topic, each input reinforces the others. If they diverge, the page is sending mixed relevance evidence.
A falsifiable test: export the queries a page earns impressions for in GSC, the anchor text of every internal link pointing to it from a crawl, and the page's main body terms. If the dominant anchor text and the dominant query terms do not appear in the body, relevance is being claimed by one input and not supported by the others.
You cannot measure T* itself. The check below measures agreement between the inputs you control, which is a proxy.
What are the common myths about T*?
**Myth: T* is keyword density.** The exhibit says body terms are one of three inputs. Repeating a phrase does nothing for anchors or clicks.
**Myth: T* is the whole ranking.** The same exhibit separates topicality from quality (see Q*) and from LLM based signals such as RankEmbed and DeepRank.
What are the key dates for T* topicality?
- 2025-01-31: DOJ call with Google engineer Pandu Nayak, later filed as exhibit PXR0357 (source)
- 2025-02-18: DOJ call with Google engineer HJ Kim describing T* and the ABC signals, later filed as exhibit PXR0356 (source)
- 2025-05-13: Search Engine Land publishes analysis of the ABC signals exhibits (source)
