GEO strategy
GEO strategy that starts with measurement, not guesswork
A GEO strategy decides where a brand should be visible in AI answers and what to fix first. Because AI assistants rarely give the same answer twice, that decision has to rest on repeated measurement rather than a single screenshot. We measure how often lawful adult brands appear in AI answers, compare them with competitors, and turn the result into priorities your team can act on.
Share of answers, repeated samples
- Brand A62%
- Brand B44%
- Brand C28%
- Your brand12%
Four instruments of a GEO strategy
- I-01Question setThe real questions your customers ask AI assistants, grouped by intent
- I-02BaselineEach question sampled repeatedly per assistant, so visibility is a range, not a guess
- I-03DiagnosisWhere you are absent, where competitors appear, and which sources the answers lean on
- I-04PrioritiesA short, ranked plan for your team or implementation partner, then a re-measure
The baseline itself is described on the LLM visibility audit page.
Why one answer proves nothing
The clearest public evidence so far comes from research run with hundreds of volunteers.
| Point | Finding |
|---|---|
| Who | Rand Fishkin (SparkToro) and Patrick O’Donnell (Gumshoe.ai) |
| How | 600 volunteers ran 12 prompts 2,961 times across ChatGPT, Claude and Google AI, Nov to Dec 2025 |
| Same brand list twice | Under 1 in 100 for ChatGPT and Google AI |
| Same list, same order | Under 1 in 1,000 |
| What held steadier | How often a brand appeared across runs (visibility), not where it ranked |
| Caveat | Not peer reviewed; the authors note they are not data scientists |
As reported by Search Engine Journal and Search Engine Land. How we turn visibility into a share-of-voice figure is set out on AI share of voice.
Outside our method
- Rank positions reported as if AI answers were fixed
- Promises of specific mentions or citations
- Tactics that try to manipulate AI answers
- Escort or sexual-services brands
Levers that tend to move answers
| Lever | How it reaches an AI answer |
|---|---|
| Clear, factual pages on your own site | Gives assistants accurate text to quote when they search the web |
| Presence on the sites assistants cite | Lists, reviews and forums in your category often supply the answer |
| Consistent facts everywhere | Name, products, prices and policies that match across sources reduce errors |
| Safe-for-work public pages | Assistants avoid explicit material, so open pages carry the visibility |
| Fresh, dated information | Assistants that browse favour pages that are current and say so |
None of these is certain to move a particular assistant; they are the levers with the most consistent evidence. We test which matter for your category by measuring before and after.
What happens if assistants open up adult topics
OpenAI announced in October 2025 that ChatGPT would allow erotica for verified adults, then delayed the feature to early 2026 and, in March 2026, delayed it again without a new date. Other assistants have their own policies. If assistants relax their rules for verified adults, answers about adult products and services could change quickly, and so could which sources they cite. A strategy built on accurate, open, safe-for-work pages and a solid presence on trusted third-party sites holds up either way. We track each assistant's policy and re-run the question set when it changes.
Building the question set
| Question group | Example (illustrative) |
|---|---|
| Discovery | Which brands make body-safe silicone toys? |
| Comparison | Brand A or Brand B for a first purchase? |
| Buying | Where can I buy discreetly in the US? |
| Trust | Is Brand A legitimate and safe to order from? |
| Problem | How do I clean a silicone toy? |
Twenty to forty questions, written the way real buyers ask, cover most categories. The set stays fixed between rounds so changes reflect the assistants, not the questions.
A quarterly GEO cycle
| Month | Work |
|---|---|
| One | Run the question set; code answers; identify the sources cited |
| Two | Fix the gaps: own-site pages, inaccurate third-party facts, missing listings |
| Three | Re-run the set; compare with the baseline; plan the next quarter |
Search Console and AI features
Google includes clicks and impressions from its AI features within the overall Search Console figures rather than reporting them as a separate line, so Search Console alone cannot show how a brand fares in AI answers. That is one good reason to measure answers directly. We combine the two: Search Console for the traffic that reaches your site, and repeated answer sampling for the visibility that may never produce a click. Looking at both together shows whether AI answers are sending people to you, mentioning you without a visit, or leaving you out.
Owned pages assistants can use
Assistants that browse the web need pages they can read and trust. For an adult brand, that usually means a clear about page, product and category pages with precise facts, guides that answer buyers' questions, and policy pages for delivery, returns and privacy, all safe for work and dated. Explicit material behind age checks is invisible to assistants, and it should stay that way for legal reasons too. Many brands find that improving five or six such pages does more for AI visibility than any other single change, and the same pages help human visitors decide too.
Who the strategy is for
GEO strategy suits lawful adult brands that already have a website and some search presence, and want to know whether AI assistants mention them and why. It is less useful for brand-new sites, which need foundations first, and it is not a substitute for compliance work such as age verification. If an audit shows your basics need fixing before anything else, we will say so. The audit itself is described on LLM visibility audit.
Third-party facts you can correct
Assistants often repeat what other sites say about you: an outdated review, a directory listing with old prices, a forum post with wrong information. Some of these can be fixed by asking the site owner, updating your own listings, or publishing a clear correction on your site that assistants can find. Keep a list of the third-party pages that appear most often as sources in your category, check them each quarter, and note which corrections were accepted.
Next reads
The difference between GEO, AEO and LLMO is explained in our definitions note. Share calculations are on AI share of voice, rivals on competitive AI visibility, and the sources assistants cite in citation source analysis.
The short version
Measure first, fix the facts and pages assistants rely on, and measure again with the same questions. Repeat each quarter, keep the question set stable, and judge progress over several rounds rather than one.
What the strategy reports
| Figure | What it shows |
|---|---|
| Share of answers naming you | Visibility across the question set |
| Share of answers citing your site | Whether your own pages are trusted sources |
| Accuracy of what assistants say | Errors about products, prices or policies |
| Change since the last round | Whether the work is moving anything |
Every figure comes with its sample size. Fees are on pricing, and the method in our guide to measuring AI visibility.
Questions
Why not track our "rank" in ChatGPT?
Because it barely holds still. In SparkToro and Gumshoe.ai research, the same prompt almost never returned the same brand list twice. How often a brand appears across many runs is a steadier measure.
Do you carry out the optimization work as well?
No. We measure, diagnose and set priorities. Your team or an implementation partner does the content and technical work, and we measure again afterwards.
Which AI assistants do you measure?
Usually ChatGPT, Google AI Overviews and AI Mode, Perplexity, Claude and Gemini, chosen by where your customers are likely to ask.
Is a GEO strategy different for adult brands?
Yes. Assistants often limit answers on adult topics, so absence can reflect policy rather than weak visibility. We separate the two before drawing conclusions.
Has ChatGPT launched an adult mode?
As of our last check, no. OpenAI announced it in October 2025, delayed it to early 2026 and, in March 2026, delayed it again without a new date. We re-run question sets whenever an assistant's policy changes.
LLM visibility audit
Find out how often AI assistants name your brand
Tell us your brand, your market and your main competitors. We run a sample of real customer questions through the major AI assistants several times, record mentions and citations, and send back a readout showing the ranges observed. The first one costs nothing and binds you to nothing.
Request an LLM visibility audit