Rankings Are Now One Column: Tracking Brand Visibility Across Google and AI Assistants

For twenty years, one number told you how search was going: your rank. Track your keywords, watch the positions, celebrate page one. That number still matters — and it is now one column in a wider table.

The buyer who used to type "fractional CMO software" into Google and click three blue links now often asks ChatGPT, or Gemini, or Perplexity — or gets Google's own AI Overview answering above the results. In each of those experiences there is no position 1 through 10. There is an answer, and your brand is either in it or it is not. Being named in the answer, and being the source the answer cites, are the new ranks. They are volatile, personalized, and they do not always agree with your organic position: a page that ranks #3 can be invisible to every AI assistant, and a brand that ranks #14 can be the first recommendation ChatGPT makes.

Why you cannot just read this off a dashboard yet

The platforms know this is a measurement problem, and 2026 brought real movement: Google Analytics added a dedicated channel for traffic arriving from AI assistants, Search Console added a Generative AI performance report showing impressions from AI Overviews and AI Mode, and Bing Webmaster Tools launched AI Performance reporting with citations and the retrieval queries behind them.

Useful — and still not enough, for three reasons. First, most of the new AI reporting is visible only in the platforms' own dashboards; the data largely is not in their APIs yet, so you cannot pull it into your own reporting or join it with your content plan. Second, each surface reports only itself: nothing shows you ChatGPT, Gemini, Perplexity, Google, and Bing side by side for the queries you actually care about. Third, and most important: none of it tells you about your competitors. "You got 400 AI impressions" means little without knowing whether the answer recommends you or the other vendor.

The teams treating this seriously do what quant investors do with market data: measure it themselves, consistently, on a schedule, and own the history. Ask the assistants your buyers' questions every week. Snapshot the search results every week. Parse who is named, who is cited, at what position. Store every measurement, because the trend — not any single reading — is the signal. A single probe of an AI assistant is an anecdote; fifty-two weekly probes of the same question are a share-of-voice curve.

What a real visibility tracker measures

If you build this yourself — and you can, with a SERP data API and grounded calls to the assistant APIs — the weekly record for each target query should carry four things:

Your Google position, measured, not estimated. Search Console gives you the true average position for queries you already get impressions on. A SERP snapshot fills in the rest: queries you do not rank for yet, and — because a snapshot returns the whole results page — every competitor's position for free, plus whether an AI Overview appeared and whose content it drew from.

Whether each AI assistant names you. Ask the buyer's question, in the buyer's words, against each assistant with web search enabled. Parse the vendors named in the answer. Record: were you mentioned, at what rank among the named vendors, and which competitors appeared. This is share of voice for the AI era.

Whether the answer cites your site. Mentions and citations are different assets. A mention means the model knows you belong on the shortlist; a citation means your content was the evidence. Citations are what your content program can most directly move, so track them separately.

What it cost you. Grounded probes and SERP snapshots have real unit prices. If you cannot see the spend per client or per business unit, you cannot make the sensible calls about cadence and coverage. Measurement without cost accounting becomes the first thing finance kills.

One honesty rule ties it together: never confuse "we did not measure" with "we measured zero." An assistant that was down last Tuesday did not stop recommending you — you stopped asking. Store the difference.

How T2D3 OS does this now

We shipped this tracker into the Content Studio's Targets tab this week, and the design choices are worth sharing because they are the same choices we would argue for in any stack.

The history lives in your workspace, not a vendor's. Every week, for each client organization's top-priority target queries, T2D3 OS appends dated measurements: Google position from Search Console and a fresh SERP snapshot, per-assistant mention and rank across Claude, ChatGPT, Gemini, and Perplexity, own-site citation share, and competitor positions — all in the same table your content plan already reads. The Targets tab shows each query's current numbers with week-over-week deltas, and expands into trend lines, a per-assistant grid, and a you-versus-competitors snapshot of the live Google top 20.

Competitors come from your foundation, not a keyword tool. The tracker measures the competitor list you curated in your market signal — the same one your positioning work uses — so "share of voice" means share against the vendors your buyers actually weigh you against.

It closes the loop into the work. This is the part a standalone rank tracker can never do. Every target query in T2D3 OS is linked to the content aimed at it — the drafts, the published pieces, the refresh queue. When a target's citation share rises after a piece goes live, you see the connection in one row. When a won query decays, the system files the refresh work itself. Measurement that does not change what gets written next week is just a prettier report.

The AI probes are glass-box and budget-capped. Every probe stores the assistant's actual answer, so you can read why you were or were not named, and every run logs its real cost per organization in the admin cost dashboard — at current prices, single-digit cents per query per week. Weekly visibility across four assistants and the live SERP costs a few dollars per month per client. The excuse for not measuring is gone.

And it is honest about what an API probe is. An assistant's API answer is not a screenshot of any particular user's session — personalization is reshaping what individual buyers see. Treat the weekly probe as a directional index, sampled the same way every week, and let the trend carry the meaning.

Rankings still matter. But the scoreboard grew columns, the platforms will not hand you the full table, and the history is the asset. Start storing it now — every week you wait is a week of trend line you can never get back.