SEO and GEO consulting for the answers your buyers actually read
Visibility in the blue links and inside the AI answer above them, measured prompt by prompt. Delivered through Geology, the AI search agency I co-founded.
Book an intro call with GeologySearch a category your company sells into and look at what Google returns. Most commercial queries now open with an AI Overview: a few confident sentences with a row of source links underneath. Below that sits the result ranked number one. Check whether that number-one result is one of the sources the Overview cited. Often it is not.
This is the gap SEO and GEO consulting closes. A brand can do everything the SEO playbook asks, win the ranking, and still be missing from the paragraph the buyer actually reads. I co-founded Geology to work on exactly this problem, and this engagement runs through Geology’s team.
Why ranking alone stopped being enough
Buyers still use Google. They also read the AI summary at the top of it, ask ChatGPT for a shortlist, and check Reddit before they decide. Each of those steps either names you or doesn’t.
The numbers on what this does to clicks are consistent across independent studies:
- Pew Research tracked real browsing behaviour and found that people clicked a traditional result on 8% of visits when an AI summary appeared, against 15% when none did.
- Ahrefs compared 300,000 keywords and found that the click-through rate for the first organic position is about 58% lower when an AI Overview is present, up from its earlier estimate.
- Google reported at I/O 2026 that AI Overviews had reached 2.5 billion monthly users, and AI Mode had passed 1 billion.
- Gartner had predicted a 25% drop in traditional search engine volume by 2026 as people moved to chatbots and agents.
Ranking first and being cited by the model are now two separate problems, and each needs its own measurement.
What the models actually use
AI answers do not simply restate the top ten results. Three patterns show up again and again in the work.
The model shortlists before it searches. On comparison-style questions, ChatGPT often settles on a handful of brands from its own training knowledge and only then goes looking for material to support them. A brand the model has no prior picture of starts the race behind, however well its pages rank. I wrote about what persistent memory in ChatGPT Atlas does to this: once the assistant remembers a user’s past preferences, the shortlist stops being the same for everyone.
Third-party sources carry the decision. Geology’s study of 83 ChatGPT shopping conversations found that user-generated content made up 38% of the sources at the comparison stage and 39% at validation, the stage where buyers want to hear from other customers. Products rated below 4.0 made up less than half a percent of recommendations. Reviews, Reddit threads and listicles are inputs to the answer, and most brands do not manage them as such.
Structure and evidence get cited. The original academic work on the subject, GEO: Generative Engine Optimization from researchers at Princeton and collaborators, published at KDD 2024, found that targeted content changes could raise a source’s visibility in generated answers by up to 40%. Pages that state facts plainly, cite their sources and carry clean structured data are easier for a model to lift.
What the engagement covers
The work runs as one programme across four disciplines, because fixing any one of them in isolation tends to stall on the others.
1. Prompt and query research
The starting point is the list of questions your buyers actually ask: typically 50 to 200 prompts for a category, drawn from real buyer language as well as keyword-tool estimates. Alongside them sit the Google queries that still drive revenue. For each prompt the research records who gets cited today, which sources the models draw on in your niche, and where competitors appear and you don’t.
2. Technical SEO and AI readability
The foundation both channels share:
- Crawl and index health. Canonicals, sitemaps, robots rules, render path for JavaScript-heavy sites, status-code anomalies and soft 404s.
- AI crawler access. Whether GPTBot, ClaudeBot, PerplexityBot and Google’s crawlers can reach the pages you want cited, and an llms.txt file where it helps.
- Schema and entity signals. Organization, Product, Article, FAQ, Author and Breadcrumb markup, plus consistent facts about the company across every page that describes it.
- Information architecture. Internal links, topic clusters, orphaned high-intent pages and cannibalisation between near-duplicate URLs.
- Performance. Core Web Vitals measured on real devices, prioritised by the revenue pages they slow down.
3. Source pages and content
The pages a model should cite for each priority prompt, built or rewritten so they answer the question directly, carry evidence and are easy to quote. That covers pillar and cluster structure, comparison and category pages, page-level rewrites and the editorial work to keep them current. Freshness matters here: in the shopping study, brand-site content cited by ChatGPT had a median age of 322 days, against 101 days for editorial sources.
4. Off-site presence
Where the model looks when your own domain is not enough: Reddit and community threads, review platforms, listicles, industry publications, digital PR and directory listings. This is the part classic SEO programmes most often leave out, and for comparison and validation prompts it is often the deciding input.
Measurement
One view of both channels: Google search traffic from Search Console, AI referral traffic from Analytics, prompt-level positions across ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews, the sources cited for each, and the sentiment of each mention. It is reviewed weekly, so the priority list moves with what is actually changing.
What you walk away with
An audit that ends in a fix list. Each finding has an owner (your engineering team, your content team or Geology), a week and an expected impact, with tickets ready to paste into your tracker. If you only want the diagnosis, the audit stands on its own.
A prompt-level visibility baseline. Where you are named, where you are missing, and who is named instead, for the questions that lead to revenue in your category.
Ongoing execution if you want it. Content, technical fixes, off-site placements and weekly tracking, reviewed monthly and reprioritised against what moved. There is no long retainer lock-in.
Results on Geology engagements
From Geology’s published case studies:
- A US specialty insurance broker went from 8,032 to 311,102 monthly search impressions and from 23 to 2,359 monthly clicks between March and August, and was named in 30 of 43 tracked non-branded AI answers. Read the insurance case study.
- A global travel insurance group grew organic clicks 112% in one quarter, with AI-answer brand coverage up 3.8 times. Read the travel insurance case study.
- A direct-to-consumer e-commerce brand grew organic revenue 48% year on year and became the default ChatGPT answer for 6 of 10 tracked buyer questions.
Insurance and financial services are where most of this work sits today. These are categories where AI answers carry real weight with buyers and where the cost of being misrepresented is high. The same programme runs for B2B SaaS, e-commerce and enterprise brands.
Who this fits
Marketing and growth leaders whose organic numbers have flattened while rankings held, founders who have noticed competitors named in ChatGPT answers where they are not, and teams whose last SEO audit never mentioned AI search. If your leadership team is working through AI more broadly than search, that is a different engagement: start with AI strategy consulting.
Not sure yet? Geology’s free live audit takes about a minute, needs no signup, and shows which AI answers name your brand against your competitors.
Questions
Are GEO, AEO, AI SEO and LLM SEO different things? +
No. They are four names for the same work: making a brand show up, and get cited, inside answers generated by ChatGPT, Perplexity, Gemini, Copilot and Google's AI Overviews. GEO (generative engine optimization) and AEO (answer engine optimization) are the two most common labels. SEO is the older discipline of ranking in the list of links, and most of its technical groundwork still applies.
Does GEO replace SEO? +
No. Google still sends most of the traffic, and AI systems lean heavily on pages that are crawlable, well structured and already trusted by search engines. GEO adds what classic SEO leaves out: which prompts matter, which sources the models cite for them, and the off-site presence on Reddit, review sites and industry publications that those models read.
How long before anything changes? +
On Geology's engagements, first new mentions in AI answers typically appear within two to four weeks, and material shifts in share of voice take eight to fourteen weeks. Technical fixes such as schema, metadata and crawl access usually go live in the first month. Timing depends on how crowded the category is and how much content debt has to be unwound first.
Can you guarantee the top spot in ChatGPT? +
No, and anyone who does is guessing. AI answers are non-deterministic and the platforms change their sourcing without notice. What the engagement commits to is the inputs (prompts mapped, pages fixed, citations earned) and weekly reporting on the outputs, so you can see what moved.
How do you measure AI visibility? +
Prompt by prompt. For each buyer question in your category, the tracking records whether your brand is named, which of your pages are cited, which competitors appear, and the sentiment of the mention. That sits alongside Google Search Console, Google Analytics and AI referral traffic, so search and AI visibility are read in one place.
What if hardly anyone in my industry searches with AI yet? +
Then the work is positioning: getting into the sources the models draw on before the category catches up, which is cheaper than displacing an incumbent later. If a category is not ready for it, the scoping call says so up front.
What does it cost? +
Scope sets the price, so there is no rate card. A single section or subdomain is a different job from a multi-site, multi-language programme. The intro call with Geology produces a written scope.
Find out where you show up today
Book an intro call with the Geology team, or run the free live audit first to see which AI answers name you and which name your competitors.
Book an intro call with Geology