How Long Does GEO Take to Show Results?
We cannot fully answer this, because we cannot fully measure the outcome. What we can do is tell you what moves in days, what takes months and what evidence would show it working, which is more useful than a number we would have to invent.
What We Can Actually Say
Some things change within days. Content and reference work accumulates across two or three quarters. Beyond that, any figure anybody gives you is invented, because the outcome cannot be measured precisely enough to time it.
Why this question is harder here than for ordinary search. The measurement problem.
For ordinary search we can tell you roughly how long because positions and traffic are measurable, so a supplier has real evidence about typical timescales. Here there is no position to track and no universal report of citations, so nobody has that evidence. Our measurement guide sets out why.
What that means for a promised timescale. It cannot have been derived.
If a supplier tells you three months, ask what they measured to arrive at it. There is no dataset of AI citation timings, so the figure came from somewhere other than evidence.
What we will commit to instead. Sequence rather than dates.
The order in which things happen is knowable. We can commit to it. Blocks two and three set out which items are quick and which are slow. Block five gives the order you should expect evidence to appear in.
Why sequence is more useful than a date anyway. It tells you when to worry.
A date gives you one moment to be disappointed at. A sequence lets you check at any point whether the thing that should have happened by now has happened, which is a better instrument for holding a supplier to account.
This page is reviewed quarterly. Along with the rest of the reasoning pages.
What Moves Quickly
Four things, all verifiable rather than inferred. If a supplier has been engaged for a month and none of these has been done, that is a reasonable thing to ask about.
Crawler access. The fastest item. The one with documented consequences.
Whether the crawlers belonging to these products can reach your site is checkable today and correctable within a day. Vendors publish reflection times for changes: OpenAI's crawler documentation, checked on 30 July 2026, notes it can take around twenty four hours for a change to a site's instructions to be reflected for search. Perplexity's documentation, checked on 30 July 2026, states the same period.
Obvious technical faults. Pages that cannot be reached at all.
Broken addresses, blocked sections, a firewall refusing legitimate requests. Days rather than weeks, then visible in server logs once fixed.
Profile and entity corrections. Wrong address, old trading name, expired service list.
The edits take an afternoon. How quickly they propagate across third party listings varies considerably, so the work is fast while the effect is not entirely.
Structured data accuracy. Making markup match the visible page.
Quick to implement. We do not present it as a lever, since no vendor has confirmed it is used in generating answers, so treat this as removing ambiguity rather than as producing a result.
What Takes Months
The three items that actually determine whether any of this works are the three slow ones. That is the uncomfortable arithmetic of the subject.
Content restructuring at scale. The largest piece of work for most sites.
Rewriting pages so passages stand alone is per page work. A dozen pages is quick. A substantial site is months. Doing it badly at speed produces material that is worse for readers, which is a real cost against a speculative gain.
Third party references. Earned rather than acquired.
Independent sources describing your business accumulate over quarters. There is no way to compress this that we would recommend, since the fast methods are the ones our backlinks guides have always argued against.
Any change in how often you are drawn on. The outcome itself.
Even where everything above is done, a shift in how frequently your material is used as a source appears slowly and unevenly. It is the hardest thing to attribute. Our sampling looks for direction across three or more rounds rather than change between two.
Why the slow items matter more. They are the ones that compound.
The quick wins in block two are mostly removing obstacles. The slow work in this block is the only part that builds anything, which is why a business unable to sustain a year should read our page on whether this is worth it before spending.
Why Model Updates Complicate It
The systems underneath this change on schedules nobody outside the vendors publishes in advance. That makes attribution genuinely difficult rather than rhetorically difficult. It cuts in both directions.
What can happen without you doing anything. Movement in either direction.
A business can start appearing in answers it never appeared in. It can equally stop appearing in ones it held, with nothing having changed on its website. The cause was on their side and it is not observable from yours.
Why that is a problem for suppliers specifically. It permits false credit.
Any improvement following a model update can be claimed as the result of the work. Any decline can be blamed on the update. A supplier with both excuses available has no accountability at all unless they commit in advance to what they would expect.
How we handle it. By writing down the expectation first.
The sequence in block five is recorded at the start, so a claim of progress can be checked against what we said would happen rather than against what happened to occur. That is the only defence against both errors.
What we will not do. Claim a change we cannot trace to work.
Where something improves and we cannot connect it to something we did, we say so. That costs us the occasional easy win and it is what makes the reporting worth reading.
The practical consequence for you. Judge across quarters.
Single month movements in this subject are frequently somebody else's release note rather than anybody's performance.
What Progress Looks Like
Four stages. They appear in this sequence rather than together. Knowing the order is what lets you judge month three without waiting for month twelve.
First, crawler activity in your logs. Weeks one to four.
Requests from the relevant crawlers arriving and being served. This is hard evidence of retrieval eligibility rather than inference. If it has not happened nothing downstream will.
Second, appearance in your sampled questions. Months two to six, unevenly.
Your business or your pages showing up in the fixed question set. Expect this to be erratic rather than a clean rise, because answers vary between askings.
Third, movement in branded search. Months four onward.
People searching your name more often, which suggests recognition is building. It is a proxy and it cannot be attributed to this work alone.
Fourth, enquiries. The only stage that pays anybody.
Last, weakest in attribution and the one that matters. Nobody can prove a given enquiry came from being cited.
Why the order matters for judging month three. It tells you what should exist by then.
At month three you should have stage one comfortably and the beginnings of stage two. If stage one is still missing, something is wrong and no amount of patience fixes it.
What Would Show It Is Not Working
The counterpart to block five. The more useful half. A supplier who cannot tell you what failure looks like has not committed to anything.
Crawler access still absent after a month. The clearest failure.
This is documented, checkable and correctable. If the logs show nothing arriving after four weeks, either it has not been looked at or it has not been fixed. Neither is acceptable.
No content restructured by month three. The second clearest.
This is the largest piece of work and the one that compounds. If nothing has been rewritten by the end of a quarter, the plan is not being executed whatever the reporting says.
Sampling flat across three quarters. Genuine evidence of a problem.
Erratic is expected. Entirely absent from your own category across nine months, while competitors appear, is a signal rather than noise.
Reporting that never mentions a limitation. A different kind of failure.
Every measure available here is a proxy or a sample. Reporting that presents any of it as certainty is either careless or managing you. Both are reasons to ask harder questions.
What we would do about it. Say so first.
If we reach month six with stage one done and nothing else moving, we would rather tell you the money is better spent elsewhere than continue billing it. Our cost page and our approach page both carry that commitment.
Sequence,
rather than dates.
There is no dataset of AI citation timings, so a promised timescale cannot have been derived from anything. What can be committed to is the order in which evidence appears, which lets you check at any point rather than waiting to be disappointed.
What is included every month:
£350 per month, one target area. No setup fee, nothing billed separately.
Twenty guides.
One subject.
The mechanism, all four platforms, what you can influence, whether it is worth it yet, what it costs, how to measure it and how to judge a supplier selling it.
