The Citation Clock

In May, Profound published a number that made its rounds through the AI search community: median time from publishing a page to being cited by ChatGPT or Claude: 6.81 days. ~900 pages, billions of LLM response logs.

Half of new pages, cited inside a week. This finding rewrites the timing assumptions every B2B content operation runs on. Most marketing orgs plan against a ranking clock that runs in quarters. The AI response citation clock runs in days.

The public clock

The full distribution is more useful than the median alone. Three-quarters of pages cited within 18.68 days. Ninety percent within 37.10. Past day 37, patience stops being a useful explanation. Pages still uncited at that point are usually held up by something more time won't solve. Profound's framing of the threshold is blunt about it: the problem is technical. (We'll come back to that.)

Two boundary conditions hold the claim in place. The study measures citation in agent logs across two engines, conspicuously omitting Perplexity, Gemini, or Google's AI Overviews. And it measures first citation, the moment a model first pulls the page, which is a lower bar than being cited consistently or recommended in the response.

So this is the clock for getting noticed, not the clock for winning. Getting noticed in a week is still a different universe from the ranking timeline, and that contrast is where the piece actually starts.

The other clock

Ahrefs' 2025 study of over a million pages is the cleanest counterweight. The average page sitting in Google's top 10 is more than two years old. The average page at position one is around five years old, up from two in 2017. When Ahrefs tracked newly published pages directly, only 1.74% reached the top 10 within a year. The rest never made it inside that window.

That's the top 10. The part of the SERP that drives B2B pipeline is the top two or three, and that real estate is older and harder still. On competitive, high-intent buyer queries, those positions are held by entrenched incumbents and aggregators. They are effectively closed to a new page on any timeline a CMO would accept.

A fair caveat: page age is a proxy for time-to-rank, not a stopwatch, and an established high-authority domain can rank quickly on low-competition terms. The contrast lives at a specific altitude. The real estate that actually matters, the top of the SERP on the queries your buyers run, takes years to reach when it's reachable at all.

The mechanism explains the gap. Search rewards accumulated trust: backlinks, engagement, domain history, signals that compound across years. AI retrieval rewards freshness and access.

A system that prefers recent, accessible, specific content will cite a new page quickly almost by design. A system that rewards accumulated authority will not. The speed difference falls out of how the two systems work, which is why it's durable.

To broadly summarize: you can influence AI answers on high-intent buyer queries faster, and against incumbency you could never crack in traditional search, than the SERP top three will ever open up.

What it looks like inside one company

The aggregate curve is one thing. Watching it happen inside a single brand is another.

We ran a 28-day engagement with a B2B human resource information systems (HRIS) platform. At baseline, the brand surfaced on 2 of 150 buyer-research queries. 1.3% visibility across ChatGPT, Claude, Perplexity, and Gemini. Effectively invisible.

The site was technically open to crawlers. The diagnosis was the most common fatal problem in GEO and the one practitioners rarely lead with. The platform was built on a client-side-rendered framework, which means AI crawlers were requesting pages and receiving empty HTML shells. GPTBot, ClaudeBot, PerplexityBot, all hitting JavaScript none of them execute. The content team was producing good material. The models couldn't read a word of it.

This is the case Profound's 37-day rule diagnoses cleanly: past the 90th-percentile threshold with no citation, more waiting won't help. The problem is technical. Here it was technical from the start.

The fix was an architectural workaround. Identical server-rendered HTML to bots and humans, routed through a worker, no re-platform required. Then a 48-hour content sprint. Then a structured internal-linking rebuild.

At the four-week re-audit, on the identical 150-query set with the identical surfacing definition, visibility moved from 1.3% to 8%. Same brand, same queries, more than six times the presence.

The part most case studies can't produce: every newly surfacing query traced back to a specific page we'd shipped, and the model was citing that page. Audit identifies the query, brief gets written, page goes live on the main domain, query surfaces at re-audit citing the exact page. A closed loop.

The clock, made visible

We took a second snapshot at six weeks. On the same matched 150-query set, visibility moved again, from 8.0% to 16%.

It doubled in two weeks, after taking four weeks to get from 1.3 to 8.

A curve that accelerates between week four and week six is the latency lag resolving. Pages shipped in the first 48-hour sprint were crossing the citation threshold weeks after they went live, exactly as Profound's distribution predicts: a median around a week, a long tail past day 37. The six-week number is substantially the payoff of week-zero work, surfacing as the engines got to it.

Profound's curve describes ~900 pages in aggregate. This engagement describes one brand over six weeks. They're the same phenomenon at two scales. The public distribution told us when citations should start landing and how long the tail runs. The client data shows that tail arriving on schedule, page by attributable page.

I haven't seen anyone pair the two, and the fit is close enough that the curve stops being abstract and becomes a planning tool. If your pages are well-built and accessible, expect movement inside the first week, most of it inside three weeks, and a tail still resolving past a month.

Fast to win is fast to lose

None of this means you publish a page and own a query. The freshness bias that opens the door in a week is the same reason it doesn't stay open. Profound's volatility data puts citation drift at 40 to 60% month over month. Content that earns a citation in week one can lose it by week six if something more specific and more recent arrives.

The shape of that drift matters. Ahrefs' AI Overview research found that the text of responses changes 70% of the time and cited URLs change 46% of the time, but cosine similarity between consecutive responses sits at 0.95. The model's underlying verdict on the category stays stable. The citation slot turns over anyway.

This cuts both ways. Search is slow to win and stable once won: reach the top three and you tend to hold it. AI presence moves the opposite direction. Fast to win, fast to decay.

A ranking is an asset. A citation is a position you re-earn, even when the model hasn't changed its mind about you.

So the advantage of the fast clock is speed of feedback. You learn whether a page worked in weeks instead of quarters, and you iterate against real signal at a tempo SEO never allowed. The brands that win here treat the clock as continuous: ship, measure at the next re-audit, defend what surfaced, replace what drifted. Presence behaves like a subscription: stop paying and it lapses.

Your velocity is the variable

Put the two clocks together and the strategic conclusion is uncomfortable for most content operations.

The engines will cite a good page in about a week. Most B2B content takes longer than that to clear legal review and design. For most companies, the binding constraint on citation speed is internal. The model is ready in seven days. The publishing process takes thirty.

The 48-hour sprint in the engagement above was made possible by architecture: server-rendered pages, pre-structured briefs, a deployment process that ships and verifies in a day instead of negotiating for a month. When the citation clock runs in days, publishing speed becomes a citation variable in its own right.

The prize for moving fast is real estate you can't otherwise reach. On the buyer-research queries that shape a B2B shortlist, comparison, validation, requirements, you can go from invisible to present in weeks, against incumbents who would hold the SERP top three for years. But you have to keep feeding the clock, because it keeps running whether you do or not.

GEO has been sold as a content discipline. The data says it's a publishing operations discipline. Architecture, briefs, deployment cadence, and re-audit rhythm are now the variables that move citations, more than any individual page is.