Search engine
Rankings in classic Google and Bing results. It’s the foundation: many AI engines look for sources in search indexes, so a page that’s invisible in Google rarely makes it into AI answers.
AI search visibility · GEO / AEOGenerative engine optimization
Generative engine optimization
you can measure.
Your buyers ask ChatGPT, Perplexity, and Gemini before they visit any website. I check whether those engines can see your site at all, whether they understand what you do, and who they cite instead of you. Then I fix what blocks the citation – and measure the result on the same prompts.
In short
What is generative engine optimization?
Generative engine optimization (GEO, also called AEO) is the work of making sure AI search engines – ChatGPT, Perplexity, Gemini, and Google’s AI Overviews – can find your page, understand it, and cite it in an answer. It covers crawler access, content structure, entities and structured data, off-site brand mentions, and citation tracking.
When it’s worth doing
- Customers say they found you through ChatGPT – or that ChatGPT recommended a competitor
- Google traffic is falling while rankings hold, because the answer appears above the results
- You sell a product or service people ask about in full sentences
- You publish expert content, but AI answers cite aggregators and forums
When to start with something else
If Google struggles to index your site, AI engines usually do too – start with an SEO audit. And if nobody searches for what you sell, in Google or in ChatGPT, the problem is demand, not visibility.
GEO, AEO, SEO
Three acronyms, one job
The industry uses different names and sells them as separate services. In practice, the same changes help with all three, so I run them together.
Answer engine
A direct answer to a question: AI Overviews, featured snippets, voice assistants. What matters is whether the answer can be lifted from your page in one piece.
Generative engine
ChatGPT, Perplexity, Gemini, and Claude assemble answers from several sources and cite some of them. What matters is whether you’re one of those sources and how the model describes your brand.
The difference is in measurement, not the work: SEO measures rankings and clicks, GEO and AEO measure presence and citations in answers. I measure both.
Scope
What I check and fix
Six areas, from technical access to what others write about you. The order matters – each one assumes the one before it works.
AI crawler access
robots.txt, CDN and firewall rules that can quietly block AI bots, rendering (many AI crawlers don’t execute JavaScript), and server logs that show who actually visits.
Citable content
The answer in the first sentence of a section, headings that match how customers ask, sections that make sense on their own, and tables and lists where they organize facts.
Entities and structured data
Unambiguous names for the brand, products, and authors; Organization, Person, and Product markup; consistent profiles linked through sameAs. Structured data clarifies entities – it doesn’t guarantee a citation.
Evidence and experience (E-E-A-T)
Specific numbers with sources, quotes, original data, an author with real experience, an update date. In the GEO study presented at KDD 2024, adding statistics, quotations, and sources raised a page’s visibility in generative engine answers by up to 40%.
Off-site mentions (digital PR)
Where your category gets discussed: roundups, comparisons, forums, reviews, trade media. I check whether you’re present and described correctly, and point out where a mention is worth earning – without buying mentions.
Measurement
A fixed prompt set checked across several engines, citation share (the share of answers to that prompt set that cite your site, against competitors), AI referral traffic in analytics, and AI bot visits in server logs.
How I work
From a prompt set to a second measurement
- 01
Prompt set
I collect the questions your customers ask – from sales calls, Search Console, and forums – phrased the way people type into a chat.
- 02
Baseline
I run each prompt in several engines and record who gets cited, how your brand is described, and where you’re missing.
- 03
Diagnosis
For every gap I look for the cause: no access, content that can’t be extracted, an unclear entity, or no mentions in sources the model trusts. I also check follow-up queries, because AI engines split one question into several searches (query fan-out).
- 04
Fixes
Tasks ordered by the business value of each question, each with an acceptance criterion. Your team implements; I answer questions along the way.
- 05
Re-measure
The same prompts after a few weeks, several runs per prompt – AI answers vary, so I look at the trend, not a single result.
Measurement
Citation share, before and after
The same prompt set, run across four engines, several times per prompt. The chart shows the share of answers that cite your site – before the fixes and a few weeks after.
Plus two signals that move faster than citations: AI referral traffic in analytics and the number of pages AI bots fetch in your server logs.
- Baseline
- After 8 weeks
AI referral traffic
+64%
Pages blocked to AI bots
−100%
Illustrative example – it shows the report format, not a specific client’s result.
What you get
One prompt, one cause, one test
The report doesn’t stop at “you’re not cited.” For every important prompt, I show who the engine picked instead of you, why, and what to change so the next measurement can check it.
A prompt where your page is missing
- Prompt
- “How much does an ERP implementation cost for a mid-size company?”
- Cited
- Two industry sites and a forum – your page isn’t there
- Cause
- Price ranges live only in a PDF, and the on-page answer starts in the sixth paragraph
- Change
- Price ranges in HTML, the answer in the section’s first sentence, a cost table by company size
- Sign-off
- The page serves content to AI bots without JavaScript; the answer sits in the section’s first 50 words; the prompt is in the next measurement
Illustrative example – it shows the format, not a specific client’s result.
Honestly
What can be measured, and what can’t
Access and traffic measurable
Server logs show whether AI bots fetch your pages. Analytics shows traffic from ChatGPT, Perplexity, or Gemini as referrals.
Citations partly
Answers change between queries and users. I measure on a fixed prompt set with repeated runs and report the trend, not a single screenshot.
Model selection out of control
Nobody decides which source a model picks for a given answer. You can raise the odds – you can’t guarantee them.
If someone promises a specific number of ChatGPT citations by a specific date, they’re promising something they don’t control.
Pricing
What drives the cost
I don’t sell packages, because the scope depends on how many questions matter to your sales. Four things affect the price:
Number of prompts and topics
A few dozen prompts for one service is different work from several hundred across a whole product catalog.
Markets and languages
Each language and market is its own prompt set and its own measurement – answers in English and Polish draw on different sources.
Technical state of the site
If the page needs JavaScript to show content, or the CDN blocks bots, technical work comes first.
One-off or ongoing
A diagnosis with a fix plan, or a diagnosis plus monthly or quarterly measurement.
After a short call, I send a proposal with scope, timeline, and price – before any work starts.
Related services
SEO
SEO audit
Technical health, content, links, and AI readiness – findings grouped by root cause, prioritized by business value, each with an acceptance criterion.
MoreContentContent audit
A decision for every URL: keep, update, merge, or remove – with cannibalization, decay, and a redirect map.
MoreContentSEO content strategy
What to publish next: a topic map, intent per page, the gap against competitors, and internal linking that builds authority.
MoreFrequently asked questions
How is generative engine optimization different from SEO?
SEO is about rankings in classic search results; GEO is about presence and citations in answers from ChatGPT, Perplexity, Gemini, and AI Overviews. Most of the work overlaps, because AI engines often look for sources in search indexes. The main differences are measurement and how much content structure matters.
Why is SEO the foundation of GEO?
Because AI engines rely heavily on what search engines have already found and evaluated. Google states that a page can only appear as a supporting link in AI Overviews if it’s indexed and eligible to show in Search with a snippet. A page with indexing, rendering, or structure problems has worse odds of ranking in Google – and with that, worse odds that a model ever sees it. That’s why every GEO project starts with a check of the technical SEO basics.
Will GEO replace SEO?
Not in the foreseeable future. A page that Google doesn’t index or trust rarely ends up in AI answers. GEO is a layer on top of SEO, not its successor.
How do I check whether ChatGPT cites my site?
Manually: ask the questions your customers ask and look at the sources under the answer. Reliably: set a fixed prompt set, run each prompt several times across several engines, and compare over time. Add AI referral traffic in analytics and AI bot visits in server logs.
Should I block AI bots?
It’s a business decision, but know the difference: some bots collect data for model training (for example GPTBot, or Google-Extended as a setting), and some fetch pages to answer search queries (for example OAI-SearchBot and PerplexityBot). Blocking the second group usually means no citations. I also check whether your CDN blocks them without you knowing.
Does llms.txt help?
I haven’t seen public confirmation that major AI engines use it to pick sources. It’s cheap, so I don’t advise against it, but I don’t start with it – access, content structure, and entities have more impact.
How long until I see results?
Access changes show up in server logs right after they ship. Citations move more slowly – engines refresh sources at different speeds, so I run the first meaningful re-measurement after a few weeks.
Do you guarantee ChatGPT citations?
No. Nobody controls which sources a model picks. What I guarantee is a data-based diagnosis, fixes with acceptance criteria, and an honest before-and-after measurement.
Let’s see who AI recommends
instead of you.
Tell me what you sell and what your customers ask. I reply within one business day.
Ask about AI visibility