The honest answer is that nothing much happens for the first quarter, and that is where most programmes get abandoned.
This is the question every prospect asks and almost nobody answers with numbers. Here are four of our published engagements with the elapsed time attached, including the flat opening.
The four
| Client type | Query set | Start | Result | Elapsed |
|---|---|---|---|---|
| IFA practice | 45 Consumer Duty queries | 0% | 31% | 90 days |
| Regulated B2B firm | 30 priority queries | 0% | 37% | 90 days |
| Regulated B2B firm, continued | same 30 | 0% | 58% | 6 months |
| Commercial law firm | 75 priority queries | 4% | 41% | 12 months |
| B2B SaaS | 180 UK buyer queries | 9% | 38% | 14 months |
Two things to notice before the percentages do any persuading.
The query set size is doing as much work as the percentage. 31% of 45 specific Consumer Duty questions and 31% of 1,000 generic category terms are completely different claims. Anyone quoting you a citation share without the denominator is quoting you nothing.
Two of these started from zero and two did not. The law firm began at 4% and the SaaS company at 9%. Starting from something is easier than starting from nothing, and the honest comparison is the percentage-point gain over elapsed months, not the headline.
The quarter where nothing happens
The law firm engagement is the useful one, because we have the quarterly trajectory:
| Month | Citation share |
|---|---|
| 0 | 4% |
| 4 | 12% |
| 8 | 26% |
| 12 | 41% |
The first four months produced an eight-point move. The last four produced fifteen.
That shape is not an accident and it is not underperformance. The opening months go on entity foundations and building an editorial workflow, and neither of those produces a citation by itself. You are making the brand resolvable and the content extractable. The citations arrive afterwards, when third-party corroboration starts accumulating on top of a brand a model can now identify.
The practical consequence: most programmes are abandoned in month three, roughly one month before the inflection. If you are buying this, budget for the flat quarter or do not start.
What actually happens at 30, 60 and 90 days
Days 1 to 30. Nothing visible. Entity work: Organisation schema, canonical naming, sameAs
links, fixing whatever disagrees about you across the platforms a model reads. Technical access:
server-side rendering, robots, llms.txt. Content restructuring on the pages that already have
something worth extracting.
Nobody is cited at day 30 who was not cited at day 1. Any agency showing you a win here is showing you a ranking, not a citation.
Days 30 to 60. The first citations, on the narrow stuff. Specific, low-competition, highly-qualified questions. On the IFA engagement the early wins were the most specific Consumer Duty questions, where no large firm had published a clean extractable answer.
Days 60 to 90. The curve starts. Two of our engagements reached 31% and 37% on their focused query sets inside 90 days, both from zero. Both were firms with genuine existing expertise that had never been structured for extraction. That matters: we were not creating authority, we were making existing authority legible.
Months 3 to 12. Compounding. 37% to 58% by month six on one engagement. 12% to 41% between months four and twelve on another. This is where the off-site work pays, and it is the part that cannot be rushed because it depends on other people publishing.
Strong SEO does not make it faster
The most counterintuitive result in the set: one firm arrived with fifteen years of reputation, page-one rankings on its key terms, and zero AI presence.
The diagnosis was strong SEO and weak entity signals. Models could not resolve who the firm was, so rankings bought it nothing. Domain Authority correlates negatively with AI visibility across the studies we have reviewed, between -0.08 and -0.21.
Existing SEO does help in one specific way: you already have content worth restructuring, which shortens the content phase. It does not transfer as authority.
What we have got wrong measuring this
We will not pretend the attribution is clean, because on our own site it was not.
On 27 September we added in-body links to all ten of our statistics pages at once. When the numbers moved three weeks later we had no untreated group left to compare against, so we could not say whether the links caused it. We had a clean cohort and treated the lot.
The right method is a declared holdout: leave a matched set of pages untouched, change the rest, compare. We now write that into the engagement letter and publish what it shows, including a null result.
Worth asking any agency: what did you leave untouched, so you could tell?
If you want your own timeline
The variables that actually move it, in rough order of impact:
- Whether your entity is already resolvable. If Wikidata, your Google Business Profile and your site disagree about your name or address, everything else waits.
- Whether you have genuine expertise already published. Restructuring beats creating, by months.
- How specific your query set is. Narrow and specific moves in a quarter. Generic head terms take a year and may not come at all without a domain authority you do not have.
- Whether anyone else will corroborate you. This is the slow one and the one outside anybody’s control.
The free AI Visibility Audit gives you your current position across five platforms for five commercial queries, which is the baseline any timeline has to be measured from.
Related
Part of a series where we publish our own numbers including the unflattering ones: 6,482 page-one rankings that produced 218 clicks, the estate audit that measured one search engine, organic at 9% of our traffic and 86% of our leads, and every published UK GEO price we could find.
Common questions
How long does it take to get cited by AI?
Structural work can surface within weeks, because fixing schema, entity consistency and answer-first structure removes things actively blocking extraction. Citation authority typically takes three to six months, because it depends on other sites corroborating you.
Across four published engagements the fastest meaningful movement was zero to 31% across 45 queries in 90 days, and the longest ran 14 months to reach 38%.
Why does nothing happen in the first three months?
Because the first quarter is spent on entity foundations and editorial workflow, and neither produces a citation on its own.
On one 12-month engagement citation share went 4% at month zero to 12% at month four. That flat opening is normal, and it’s exactly where most programmes get abandoned, usually one month before the inflection.
What can I expect at 30, 60 and 90 days?
At 30 days, technical and entity fixes are in and nothing is visible yet. At 60 days you should see the first citations on low-competition, highly specific queries.
At 90 days a well-run programme on a site with existing expertise can reach 30% to 37% citation share on a focused query set, which is what two of our engagements did from a standing start. Head terms take considerably longer.
Does it get faster if I already rank well on Google?
Not reliably. One firm came in with fifteen years of reputation and page-one rankings on its key terms and had zero AI presence.
Domain Authority correlates negatively with AI visibility in the studies we’ve reviewed. Strong traditional SEO gives you content worth restructuring, which helps, but it doesn’t transfer as authority.
Can an agency guarantee AI citations in 30 days?
No, and treat the offer as a warning. Citation depends on third-party corroboration accumulating on sources the model already trusts, which is outside any agency’s direct control.
What can be promised in 30 days is the structural work: schema, entity consistency, answer-first rewriting. Those are real and necessary, and they aren’t citations.