How long until ChatGPT and Perplexity actually cite you?

Every guide to generative engine optimization explains what to do and none of them say when it pays off, which is the only question a business actually asks before signing anything. That gap existed for a good reason: until recently nobody had published a controlled measurement. Now somebody has — a set of pages published on one domain and checked daily against AI answer engines for a month — and the results are specific enough to plan against. They are also uncomfortable, in a way that is more useful than the optimism they replace. Click through each phase below for what happens when, what to measure while it happens, and the milestones that mark real progress rather than noise.

The first six months, week by week

Week 0 is the day the page goes live. Select a phase to see what actually happens in it, what to measure, and the milestones that fall inside it.

What each engine needs first, and how fast it reacts

None of these engines offer a submission form, a priority queue or a paid inclusion route. The only levers are being crawlable, being indexed where indexing is the gate, and being worth selecting.

Why “when” went unanswered for so long

Every other question about generative engine optimization has been answerable since the term was coined: what it is, how answers get assembled, what kind of writing gets lifted cleanly. The timing question stayed open because answering it needs something nobody had bothered to build — a set of pages published at a known moment and checked against several answer engines every day afterwards. Without that, every claimed timeline was somebody’s impression of their own client work, and impressions are exactly the wrong instrument for a system that regenerates its answers every couple of days.

The measurement now exists. Eighty-one pages, published on one domain, queried daily for thirty days across Google’s AI Mode and ChatGPT search. It is one study on one domain and it should not be treated as a universal constant — but it is the difference between a number and a vibe, and its shape is consistent enough with what we see on client sites to plan against.

What it settles, more than anything, is that “how long until AI cites us” has two different answers depending on which AI you mean. Treating the answer engines as one channel is the root of most disappointment here, because their curves are not just different in speed, they are different in shape.

The clock starts at crawl, not at publish

The first week is not a content problem, and treating it as one wastes it. Four user agents decide whether anything downstream can happen, and each operator publishes its own rules. OpenAI runs OAI-SearchBot to surface sites in ChatGPT’s search features and GPTBot for model training, both of which respect robots.txt, plus ChatGPT-User for fetches a person triggered by asking a question — and OpenAI is explicit that robots.txt rules may not apply to that last one, because a user asked for it. Perplexity splits the same way: PerplexityBot builds the index and obeys robots.txt, while Perplexity-User fetches live and generally ignores it.

Both operators quote roughly the same latency on the control side. OpenAI says it can take about 24 hours from a robots.txt update for their systems to adjust; Perplexity says it may take up to 24 hours for changes to be reflected. That is a small number with a sharp edge: an accidental block costs a day to undo even after somebody notices, and the most common cause is not robots.txt at all but a bot-protection rule in a CDN that nobody remembers switching on.

Google is a different case entirely, and simpler. Its own documentation states there are no additional requirements to appear in AI Overviews or AI Mode and no special optimisations necessary — but that to be used as a supporting link, a page must be indexed and eligible to be shown in Google Search with a snippet. So the AI clock and the classic indexing clock are the same clock. If a page is not indexed, no amount of GEO work reaches it, which is why the first twelve months of SEO after launch is the prerequisite reading rather than a competing subject.

Google moves in days. ChatGPT moves in weeks.

The gap at the start is larger than most people expect. Within 24 hours of publishing, 36% of the test pages were already cited somewhere in Google AI Mode. ChatGPT search had picked up 10% in the same window — more than three times slower off the line. By day seven, Google was at 56% and ChatGPT at 17%.

Then the curves change character. Google’s coverage fluctuated week to week, rising to a peak of 59% and falling back repeatedly. ChatGPT’s climbed steadily and did not give ground: 35% at two weeks, 42% at thirty days. Google is fast and noisy; ChatGPT is slow and cumulative.

That has a direct consequence for how you report on this work. A GEO project reviewed at four weeks looks like a Google success and a ChatGPT failure, and both readings are artefacts of the review date. The correct checkpoint is somewhere past six weeks, and the correct framing is two separate curves rather than an average that describes neither.

Perplexity sits outside this comparison for a structural reason: a significant share of its crawling is triggered by a user’s question rather than by a schedule, so its behaviour depends heavily on whether anyone is asking about your topic. On a narrow, specific prompt it can be the fastest of the three. On a broad one it can lag both.

The ceiling nobody quotes

The number worth carrying out of the study is not a speed, it is a limit. Over a full month, Google AI Mode peaked at 59% of the test pages and ChatGPT search reached 42%. That is on a domain with strong established authority publishing content squarely in its own subject area. Even under those conditions, roughly half of what was published was never cited anywhere.

This is the honest correction to how GEO is usually sold. The pitch implies that structuring content correctly makes it citable; the data says structuring content correctly makes it eligible, and selection is a separate, competitive step you do not control. The original research that defined the field found the same thing from the other direction — visibility gains of up to roughly 40% within a generated answer, and which technique produced them varied by domain. Both findings point the same way: this is a real, bounded lever, not a switch.

Plan on that basis and the economics stay sane. A page that fails to get cited is not a failed page — it is the expected outcome for about half of them, and the ones that do get cited tend to keep the citation. Publishing a small number of genuinely specific answer pages beats publishing many broad ones, because the ceiling is per-page and the broad ones lose to bigger entities every time.

A single check tells you almost nothing

This is the part that changes how you measure, and it is the least intuitive finding in the whole area. Ahrefs tracked more than 43,000 keywords with at least sixteen recorded AI Overviews each and found an average persistence of 2.15 days — an AI Overview has roughly a 70% chance of being different from one observation to the next. More usefully: only 54.5% of the cited URLs overlap between consecutive runs of the same query. Close to half the source list is replaced every time.

What does not change is the meaning. The same query re-run produces answers with an average cosine similarity of 0.95, and 54% of the named entities stay put. The system is confident about what it thinks and casual about who it credits — which means a citation appearing or disappearing on any given day carries far less signal than it feels like it does.

The practical rule follows immediately: never measure AI visibility with a single check. Run a fixed prompt list at least ten times across several days and report the share of runs in which you appear. A binary “are we cited” screenshot is roughly a coin flip, and we have watched more than one agency relationship turn on one taken at the wrong moment.

Ranking still helps. It helps less than it did.

For most of the AI-search era the reassuring answer to “how do we get cited” was “rank well, the rest follows”. In July 2025 that held up: Ahrefs analysed 1.9 million citations from a million AI Overviews and found 76.1% of cited pages ranked in Google’s top 10, with 86% appearing somewhere in the top 100.

By March 2026 the same analysis over 863,000 SERPs and four million AI Overview URLs put the top-10 share at 37.9%, with 31.2% from positions 11 to 100 and 31.0% from beyond the top 100 entirely. Ahrefs attribute the shift to Google changing the model behind AI Overviews in January 2026 and expanding its query fan-out more aggressively, pulling sources from related searches rather than the one the user typed.

Two things follow, and they pull in opposite directions on purpose. Ranking is no longer a reliable route to citation, so a strong SEO position is not the guarantee it was. But not ranking is no longer disqualifying either — roughly a third of citations now come from pages that do not appear in the first hundred results, which is the first genuinely new opportunity this field has produced for smaller sites. Where the budget should sit between the two is covered in SEO or GEO, and the mechanics of the work itself in generative engine optimization explained.

What we would tell a client to expect

Week one: confirm every crawler can reach the page and that it is indexed. If Google’s AI surfaces have not cited anything within a fortnight and the page is indexed, the problem is the content’s specificity, not the timeline.

Weeks two to six: expect Google-side movement and expect it to be jumpy. Do not report a number from a single day, do not celebrate one, and do not panic at one. Build the prompt list now and freeze it, because changing it later throws away the comparison.

Weeks six to twelve: this is when ChatGPT arrives, and when the work either justifies itself or does not. It is also the earliest point a review is worth holding.

Months three to six: the plateau. Further gains come from off-site mentions, entity consistency and freshness on time-sensitive pages, not from another pass over the same content. If the ceiling on your prompt set is low, the honest answer is usually that a bigger entity owns those questions and a narrower set of questions is the better investment.

And a caveat that belongs in every one of these conversations: Google changed the model behind AI Overviews in January 2026 and the citation mix moved substantially. Any timeline in this field, including this one, is a description of a moving system rather than a law.

What to do while you wait

The waiting period is not idle time, and the work that pays during it is unglamorous. Fix what the crawlers see: server-rendered content, clean headings, a direct answer near the top of each section rather than three paragraphs in, and structured data that states plainly who you are and what you sell. None of that is AI-specific — it is the same technical foundation classic search rewards, which is why the two disciplines are additive rather than alternatives.

Then work on the half you cannot edit. Answer engines cross-reference sources rather than trusting one, so the same facts about your business appearing consistently across directories, profiles, comparisons and press does more for attribution than another rewrite of your homepage. Inconsistency here is quietly expensive: a business described three different ways in three places gets attributed less confidently than one described identically everywhere.

And keep publishing narrowly. A page that answers one specific question completely gets cited for that question far sooner than a broad guide gets cited for anything, because a narrow page is easier to lift a clean passage from and has fewer entities competing for the same slot. GEO Starter is €599 and is our scoped version of this work — extractability, schema, an llms.txt file and an entity consistency check — with SEO Starter at €499 as the foundation it assumes underneath.

Sources

The citation timings, volatility figures, ranking overlap and crawler rules above come from these, checked August 2026. The phase structure, the measurement advice and the read on what it means for a smaller site are ours.

• Semrush — How Fast Do AI Search Platforms Cite New Content? ↗

  The controlled test behind every day-by-day figure here: 81 newly published pages on a single strong domain, tracked daily for 30 days. Google AI Mode cited 36% within 24 hours, 44% by day two, 56% by day seven, peaking at 59%. ChatGPT search reached 10%, 12%, 17%, then 35% at two weeks and 42% at thirty days.

• OpenAI — Overview of OpenAI Crawlers ↗

  OAI-SearchBot surfaces websites in ChatGPT’s search features and obeys robots.txt, with a stated latency of about 24 hours from a robots.txt update for OpenAI’s systems to adjust. GPTBot crawls for model training. ChatGPT-User handles user-initiated fetches, where robots.txt rules may not apply.

• Perplexity — Perplexity Crawlers ↗

  PerplexityBot exists to surface and link websites in Perplexity’s search results, obeys robots.txt, and takes up to 24 hours for changes to be reflected. Perplexity-User supports user-initiated actions and generally ignores robots.txt, since a person requested the fetch.

• Google Search Central — AI features and your website ↗

  Google’s own position: there are no additional requirements to appear in AI Overviews or AI Mode and no special optimisations necessary, but a page must be indexed and eligible to be shown in Search with a snippet to be used as a supporting link. AI feature performance is reported inside the Performance report under the “Web” search type.

• Ahrefs — AI Overviews Change Every 2 Days (But Never Change Their Mind) ↗

  Louise Linehan, November 2025. Over 43,000 keywords with at least 16 recorded AI Overviews each: 2.15 days average persistence, roughly a 70% chance of changing between observations, 54.5% average URL overlap between consecutive runs — but 0.95 average cosine similarity between the answers themselves, and 54% of entities unchanged.

• Ahrefs — 76% of AI Overview Citations Pull From the Top 10 ↗

  Louise Linehan, July 2025. 1.9 million citations from one million AI Overviews: 76.1% of cited pages ranked in Google’s top 10, 9.5% at positions 11–100, 14.4% nowhere in the first hundred results.

• Ahrefs — Update: 38% of AI Overview Citations Pull From The Top 10 ↗

  Louise Linehan, March 2026. The same analysis over 863,000 SERPs and 4 million AI Overview URLs: 37.9% from the top 10, 31.2% from positions 11–100, 31.0% from beyond the top 100. Attributed to AI Overviews moving to Gemini 3 in January 2026 and a more aggressive query fan-out.