Evidence on this page last checked on 24 September 2026.
When a buyer asks ChatGPT, Claude, Gemini or Perplexity which vendors to look at, the answer names a few companies and cites the pages it trusted. GEO consulting is the work of becoming one of those names, and showing it with data you can check rather than screenshots.
I have worked in SEO for 13+ years, directly with founders and marketing leads in the US, UK, Canada, Australia, UAE and India. Everything below comes from real runs: the method, the client evidence, and the prompts where my own site is not cited yet.
Who needs GEO consulting, and who doesn't?
You need it when buyers ask AI assistants for vendor shortlists in your category and competitors get named while you don't. You don't need it yet if your site barely ranks: in my September test, 34 of 40 AI answers cited live web pages, so a site search engines can't find rarely gets cited.
A good fit if
- You rank in Google for category terms, but AI answers to "best [category] tools" name competitors and cite roundup articles instead of you.
- AI engines describe your product wrongly: old pricing, missing integrations, or positioning you dropped.
- You need a number you can report upward: which buyer prompts, which engines, cited or not, month over month.
- You can ship page changes, or want them written and implemented for you.
Not yet, if
- Your site barely ranks or is not fully indexed. Fix that first; SEO consulting covers it.
- You want a guaranteed "#1 in ChatGPT". Answers vary from run to run and nobody controls them.
- You only need an llms.txt file. Google says you don't need "new machine readable files, AI text files, or markup" to appear in AI Overviews or AI Mode (Google Search Central), so on its own that is not an engagement.
What does a GEO engagement include?
Six workstreams, run in order of expected impact: an AI citation audit across four engines, a citation-gap analysis, entity and knowledge-graph cleanup, citable page rewrites, crawler access and llms.txt, and third-party presence on the pages engines already cite. Every recommendation comes with the prompt data behind it.
AI citation audit across four engines
Your buyers' real questions, taken from sales calls, Search Console queries and "People also ask", become a fixed prompt set. I run it through ChatGPT, Claude, Gemini and Perplexity with web search on and record, for every answer, whether you are cited, whether you are named, and which URLs were cited instead.
Citation-gap analysis
Who gets cited in your place, and why. For category questions it is usually third-party roundups: when I asked all four engines "Who are the best GEO consultants for B2B SaaS?", 37 of the 50 pages they cited were roundup or ranking lists, and only 2 were an agency's own service page.
Entity and knowledge-graph consistency
One version of the facts about your company wherever an engine looks: your site, Organization and product schema, LinkedIn, review sites such as G2, Crunchbase, and Wikidata where you qualify. Engines repeat whichever version they find. In my own September test, Claude described my results with a figure from an older page ($385K+) instead of the current $583,956.
Citable page rewrites
Answer-first blocks under question headings, comparison tables an engine can quote, facts with a named source and a date, and a visible last-updated date. On the Indigo Software engagement that meant one question per H2, answered in the first two sentences with the number stated, and every product, edition and version named in full.
Crawler access and llms.txt
Check that robots.txt, your CDN and your firewall let AI crawlers in, and that key content is in the HTML without JavaScript. OpenAI documents that sites opted out of OAI-SearchBot "will not be shown in ChatGPT search answers", while GPTBot is its training crawler, so each needs its own decision. llms.txt gets added as housekeeping, not as the fix.
Third-party presence
Getting onto the roundups, directories, review sites and community threads that engines already cite for your category, through editorial pitches, genuine customer reviews and useful answers. No paid placements dressed up as editorial, and no fake reviews.
Sources: OpenAI, Overview of OpenAI crawlers; roundup count from my 24 September 2026 baseline, all cited URLs classified by hand.
How do I measure AI citations?
I run a fixed set of buyer prompts through ChatGPT, Claude, Gemini and Perplexity with web search on, using the same settings every month, and log every cited URL. Then I cross-check with GA4's AI Assistant channel and Search Console. One run is a sample, not a ranking, so I report trends, not single wins.
- Buyer prompts, fixed for the engagement. Written the way people ask, for example "Recommend a freelance SEO and GEO consultant for a B2B SaaS company." Category, comparison and problem questions, agreed with you up front.
- Four engines, same settings every month. Each prompt runs through ChatGPT, Claude, Gemini and Perplexity via their APIs (I use DataForSEO's LLM Responses endpoint) with web search on and a US location where the engine accepts one.
- Everything logged. The answer text, every cited URL, whether you are named, and the searches the engine ran first. For the prompt above, ChatGPT searched for phrases like "B2B SaaS SEO consultant freelance" before answering, which tells you which searches you need to rank for.
- Redirect links resolved. Gemini returns its sources as Google redirect links, so a plain domain match counts zero citations. In my own September run, resolving all 116 of Gemini's links surfaced a citation of my site that the raw data had missed.
- Reported as AI share of voice. Per prompt: cited, named but not cited, or absent, with the URLs that won instead. Alongside it: GA4's AI Assistant channel (sessions from chatgpt.com, claude.ai, gemini.google.com and perplexity.ai) and Search Console data on Google's AI features where the property shows it.
What the tracker can't tell you. Answers change between runs, users and days. One run per prompt per month is a sample, not a rank. That is why the prompt set stays fixed and I report the trend across months, not one good answer.
What results can I show you?
Three kinds of evidence, all dated. A client's impressions inside Google's AI features grew 11-fold in four months. My own site was cited in 3 of 40 answers across 4 engines on 24 September 2026, a small sample, and in none of the answers to the GEO-consultant question. GA4 logged 114 AI-assistant sessions from June to September 23.
Client result: Google AI-feature impressions, up 11-fold in four months
My SEO engagement with Indigo Software, a US Microsoft software reseller, has run for 28 months (paused in March and April 2026) and produced $583,956 in verified organic revenue. The client's own Search Console shows impressions inside Google's AI features, AI Overviews and AI Mode, growing "on the same content, with no new link building aimed at it."
| Month | Impressions | Change |
|---|---|---|
| May 2026 | 19,543 | Baseline, first month measured |
| Jun 2026 | 51,640 | +164% |
| Jul 2026 | 54,055 | +5% |
| Aug 2026 | 215,070 | +298% in one month, +1,001% since May |
These are impressions, not clicks or citations. The same case study reports the counterpoint: Bing AI citations fell from a June peak of 36,900 to 9,000 in August. Read the GEO section of the Indigo case study.
My own site: 40 AI answers on 24 September 2026
Ten buyer prompts, four engines, web search on: 40 answers, a small sample. The site was cited in 3 of them, shown with its position among each answer's cited sources:
| Engine (model) | Prompt | Result |
|---|---|---|
| ChatGPT (gpt-5.5) | "Recommend a freelance SEO and GEO consultant for a B2B SaaS company." | Cited the home page, 2nd of 5 sources; named first among the alternatives to its top pick |
| Claude (claude-sonnet-5) | "Best SEO consultant in India for SaaS and software companies" | Cited an Insights article, 1st of 6 sources; listed first under individual consultants |
| Gemini (gemini-3.5-flash) | "Who are the best SaaS SEO consultants for B2B software companies in 2026?" | Cited an Insights article, 3rd of 10 sources, visible only after resolving Gemini's redirect links |
| All four engines | "Who are the best GEO (generative engine optimization) consultants for B2B SaaS?" | Not cited. 37 of the 50 pages cited were roundup or ranking lists |
Perplexity did not cite the site for any of the ten prompts. That is the baseline I measure my own GEO work against, in the same format you would get.
AI-assistant referrals in GA4
| Month | Sessions | By source |
|---|---|---|
| Jun 2026 | 25 | claude.ai 13, chatgpt.com 8, gemini.google.com 4 |
| Jul 2026 | 42 | claude.ai 31, chatgpt.com 11 |
| Aug 2026 | 29 | chatgpt.com 15, claude.ai 14 |
| Sep 1-23, 2026 | 18 | chatgpt.com 12, perplexity.ai 6 |
114 sessions in total: small numbers, reported as they are, and not a straight line up. ChatGPT referrals produced 5 GA4 key events (form submissions or confirmed bookings), and 22 of the sessions landed on the home page.
How does a GEO engagement run?
Four stages. First a baseline: your prompt set run across all four engines, plus GA4 and Search Console. Then a gap analysis and a ranked fix list. Then the fixes, on your site and off it. Then the same prompts re-run every month against the day-one baseline. Engagements have a 3-month minimum, then run month to month.
Baseline
Build the prompt set with you, run it across ChatGPT, Claude, Gemini and Perplexity, and pull GA4 AI Assistant sessions and Search Console data. You see exactly where you stand before any work starts.
Gap analysis
List who is cited instead of you, prompt by prompt, and why: roundups, review sites, community threads or a competitor page. Every fix is ranked by the number of prompts it affects.
Fix
Entity cleanup, page rewrites and crawler access first, then outreach to the roundups and directories engines already trust. On-site changes are implemented directly or handed to your team as exact specs.
Re-measure
Same prompts, same engines, same settings every month. You get cited, named or absent per prompt next to the baseline, plus what changed and what happens next.
GEO consulting FAQs
GEO (generative engine optimization) is the whole effort to get a brand cited in AI-generated answers, including off-site work such as review sites, directories and roundup lists. AEO (answer engine optimization) is narrower: structuring individual passages so an engine can lift them as a direct answer. You need both, and I run AEO as one workstream inside a GEO engagement.
A dated baseline of your buyer prompts across ChatGPT, Claude, Gemini and Perplexity; a citation-gap report naming the URLs cited instead of you; entity and schema fixes; rewritten pages; a crawler and llms.txt check; outreach to the roundups and directories engines already cite; and a monthly report against the baseline. If a consultant cannot show you the prompt data, ask why.
There is no fixed price list. Cost depends on how many prompts and markets you need tracked, how much of the site needs rewriting and how much off-site work your category needs. Book a free 30-minute call and you get a scoped proposal within 24 hours. Engagements have a 3-month minimum, then run month to month, with no setup fees.
Plan for months, and measure every one of them. On the Indigo Software engagement, impressions inside Google's AI features went from 19,543 in May 2026 to 54,055 in July, then 215,070 in August. Crawler and schema fixes ship quickly; entity and third-party work takes longer to show. Re-running the same prompt set monthly shows which of them is moving.
Yes. In my September test, 34 of 40 AI answers cited live web pages, and the engines searched before answering: asked for a freelance SEO and GEO consultant, ChatGPT first searched phrases like "B2B SaaS SEO consultant freelance". Pages that rank for those searches are the pool it cites from, so GEO builds on SEO rather than replacing it.
Mostly no. The groundwork is shared: crawlable pages, answer-first content, consistent entity facts and a presence on the sources engines trust. What differs is how widely each engine cites. For the same GEO-consultant question in my September test, Perplexity cited 28 different pages and ChatGPT 8, so I track each engine separately and fix the gaps that differ.
