The short answer
GEO results arrive in layers, not on one date. Technical fixes can be shipped and verified within days. Engines that search the live web, such as Perplexity, can reflect changes within weeks of your pages being re-crawled. Answers that ChatGPT, Claude and Gemini give from what the model already knows can take months, because they only change when the model or its retrieval behaviour changes.
So the honest answer to “how long does GEO take?” is: days for what you control directly, weeks for live-search engines, and months for model memory. Anyone promising a single date across all four engines is guessing.
GEO timeline at a glance
| Stage | What happens | What you can verify |
|---|---|---|
| Week 1 | Baseline measurement across engines; technical readiness fixes | Crawler access, HTML rendering and schema, checked the day they ship |
| Weeks 2–6 | Answer-first pages published or restructured; pages re-crawled | Indexing in Google and Bing; first movement in live-search engines |
| Months 2–3 | Third-party work: directories, review sites, publications, Wikidata | Citation and naming rates per engine against the baseline |
| Months 3–6+ | Models update; third-party sources are re-read | Change in answers given from model memory, engine by engine |
These are the stages the work moves through, not guaranteed outcomes. The later rows depend on engines and third-party sources you do not control, which is why each one is measured rather than promised.
Days: what you can change straight away
Technical readiness is in your control: AI crawlers allowed in robots.txt, content visible in the HTML without JavaScript, structured data that describes your organisation, a sitemap, and an accurate Wikidata record. Each of these can be checked the day it ships, for example with our free AI-readiness checker.
Fixing them removes reasons for an engine to skip you. It does not by itself make an engine recommend you. Think of it as getting through the door, not getting the recommendation.
Weeks: engines that search the live web
Perplexity searches the live web for most answers, and ChatGPT, Gemini and Google AI Overviews search for some questions too. When an engine retrieves pages at answer time, a clear, well-structured page can start being cited once it has been crawled and indexed, so changes there tend to show up first.
What speeds this up is the same thing that helps extraction: a direct answer near the top of the page, short self-contained sections, question-style headings and FAQ markup. For how retrieval engines pick which passages to cite, see how AI chooses which sources to cite.
Months: answers from model memory
ChatGPT and Claude answer some buyer questions from what the model already knows, without searching. In our payments measurements, Claude answered 5 of 11 buyer questions from memory in every run, so no website could be cited on those questions at all.
Movement on those questions depends on what the models learn over time and when they are updated, which you cannot schedule. The work that influences it is mostly off your own site: being described accurately and consistently in the publications, directories, review sites and reference sources that models learn from.
What depends on third parties
Much of what engines cite is not your own site: industry publications, directories, review sites, comparison pages and community discussions. Getting accurately represented in those sources takes as long as those sources take to publish and to be re-read by each engine. This is usually the slowest controllable part of GEO, and the part that lasts longest once done.
What to expect in your first 90 days
Month 1: a baseline across ChatGPT, Perplexity, Claude and Gemini, technical fixes shipped and verified, and a short list of the buyer questions where competitors are named and you are not.
Month 2: answer-first pages for those questions published or restructured, internal links pointing to them, and outreach started on the third-party sources engines already cite in your category.
Month 3: a second dated measurement, compared like for like with the baseline. Expect live-search engines to move before answers given from memory. If nothing has moved anywhere, that is a signal to revisit which questions and sources you are targeting.
For the full playbook on getting there, see the guide to getting cited in AI search, and for B2B software specifically, GEO for B2B SaaS.
Why GEO timelines vary so much
Three things decide how fast you move. First, where you start: a company already named in reliable third-party sources has less ground to cover. Second, the engine mix your buyers use: live-search engines respond sooner than answers drawn from memory. Third, your category: in crowded categories, engines have more established sources to choose from, so displacing them takes longer.
That is also why published “results in 30 days” claims are rarely backed by repeatable measurement. GEO is new, and engines change their models and how they search often.
How to measure progress honestly
Use a fixed set of buyer questions, ask each one several times per engine, and report cited and named separately for each engine. Compare dated snapshots like for like: the same questions, the same engines, the same method. A single good answer is not a trend, and one engine is not a proxy for the others.
What we commit to
We measure and report what changes, engine by engine, with the method published. We do not guarantee scores, rankings, citations or dates, and we say so in our terms.
Frequently asked questions
See where you stand in AI answers.
A free audit of your category across ChatGPT, Perplexity, Gemini and Claude. Delivered within 48 hours. No call required.
Get the free audit