Generative Engine Optimisation for SaaS
You will also see this called AI search optimisation, LLM SEO or AI visibility work. New to the vocabulary? What is generative engine optimisation, what is answer engine optimisation and GEO vs AEO vs SEO answer the terms before the offer.
What we can actually evidence.
How these are measured
- SaaS brands served:
- n = 30 SaaS brands. company inception to August 2026. Count of distinct SaaS companies under a signed advisory agreement. Excludes one-off audits, workshops and prospect calls.
- Years in-house experience:
- n = 1 founding team. 2015 to 2026. Cumulative years spent in salaried, in-house growth and marketing roles at SaaS companies before founding Surge45.
- AI assistants tracked:
- n = 6 assistants. current monitoring coverage. ChatGPT, Gemini, Claude, Perplexity, Microsoft Copilot and Grok, plus the AI Overviews and AI Mode surfaces inside Google Search. A coverage fact, not a performance claim.
- Growth pillars, one system:
- n = 4 pillars. current advisory model. Discover, Convert, Attribute and Grow. A structural fact about how the advisory is organised, not a performance claim.
Most of your citations will never come from your own site
An analysis of 5.7 million citation links across B2B SaaS prompts found the ten most-cited domains account for 35%+ of all citations. Reddit is first, G2 second. Eight of the ten are domains a SaaS vendor does not own.
Almost every AI search offer in the market sells schema, entity work and content restructuring. That work is necessary and it addresses the minority of citations. The coordinated programme for the sources that actually get cited is harder, slower and needs senior judgement, which is why few people sell it and why we lead with it.
So the honest version of this offer is: on-site work alone will not deliver the outcome. If that is all you want, there are cheaper places to buy it.
Source: Goodie, The Most Cited B2B SaaS Domains in AI Search. 5.7 million citation links analysed across B2B SaaS prompts.
Every assistant, and what we cover
Including the two we have deliberately chosen not to write pages about. A stated exclusion is a judgement. An unstated one is an oversight.
| Assistant | US share | Our coverage |
|---|---|---|
| ChatGPT | 51.3% | Dedicated page |
| Google Gemini | 27.7% | Dedicated page |
| Claude | 10.3% | Dedicated page |
| Grok | 2.8% | Dedicated page |
| Perplexity | 2% | Dedicated page |
| Microsoft Copilot | 1.3% | Dedicated page |
| DeepSeek | 0.4% | No page, deliberately. Under half a percent of US assistant usage and almost no SaaS buying activity. We monitor it and will publish a page if that changes. |
| Meta AI | 0.1% | No page, deliberately. Consumer-social usage rather than software research. No SaaS buyer journeys observed in our monitoring. |
Source: First Page Sage, Top Generative AI Chatbots by Market Share. July 2026, US monthly-active-user estimates.
How a GEO engagement runs
Four things cause low AI visibility and they have very different remedies, so the first job is establishing which one you have rather than proposing a programme.
- 1
Baseline
Declare and freeze the prompt set and competitor set. Run every prompt repeatedly across every surface. Record answers, sources and model versions.
- 2
Diagnose
Separate the four causes: blocked retrieval, poor extractability, weak entity clarity, or absence from the sources being cited.
- 3
Fix what is cheap and unambiguous
Crawler directives, stale third-party listings, opening passages on pages that already rank. Days and weeks of work, often the largest single movement.
- 4
Build the off-site programme
Review platforms first because they are fastest, then community, then publisher and analyst placement. Quarters of work, sequenced by where your citations actually sit.
- 5
Re-measure and report
The same frozen prompt set, per platform, with citation rate and mention rate kept separate. A change in method starts a new series rather than restating history.
- 6
Decide what to stop
Every review includes what is not working and what we would drop. A programme that only ever adds is not being managed.
Buy the whole programme, or one part of it
Answer engine optimisation
The content layer on its own: writing so a machine can lift one passage and be correct. Necessary, cheaper than the full programme, and not sufficient by itself.
Learn moreSaaS AI search advisory
Direction and judgement with your team executing. For companies with capacity but no senior person who has run this specifically.
Learn moreCitation sources programme
Review platforms, communities, analyst and publisher placement, and video. Where most citations actually come from.
Learn moreThe assistants
Products a buyer opens deliberately. Each page below is written around that platform's own retrieval mechanism: where its index comes from, which crawler fetches what, how citations are chosen and displayed, and what has actually been observed to move it.
ChatGPT
51.3%ChatGPT SEO starts here. The largest assistant by a distance, and the one whose citation behaviour changes most between a model answering from memory and a model running a search.
Read the mechanismGoogle Gemini
27.7%Gemini SEO starts here. The second-largest assistant by usage, and the one most often missed because two pages about Google Search surfaces make Google look already covered.
Read the mechanismClaude
10.3%Third by usage, but over-indexed among engineering and technical buyers, which makes its share understate its commercial weight for developer-facing SaaS.
Read the mechanismGrok
2.8%Fourth by usage, and the only assistant whose answers respond to what people are saying about you this week rather than this year.
Read the mechanismPerplexity
2%Perplexity SEO starts here. Small by usage and disproportionately useful, because it is the one assistant that shows its working: sources are listed on screen, so citation share is directly observable rather than inferred.
Read the mechanismMicrosoft Copilot
1.3%Copilot SEO starts here. Small in consumer usage, structurally important in enterprise, and the one platform where the work is mostly not AI work at all.
Read the mechanismGoogle's Search surfaces
These are separated from the assistants on purpose. AI Overviews and AI Mode are surfaces inside Google Search, shown to somebody who is searching. The Gemini app is a separate product a buyer opens deliberately. Collapsing the two is what made Google look covered on this site while the single largest gap sat inside it.
Google AI Overviews
AI Overviews optimization starts here. A generated summary above the conventional results, on the queries Google decides warrant one. Not a separate product, and not the Gemini app.
Read the mechanismGoogle AI Mode
A conversational search experience the user opts into inside Google. Same index, different interaction, and a materially different retrieval pattern.
Read the mechanismThe sources that carry the citations
Four strands, run as one programme and reported separately because they move on very different timescales.
Review platform strategy
G2, Capterra and GetApp all sit in the ten most-cited domains for B2B SaaS prompts, with G2 second only to Reddit. Assistants treat review marketplaces as structured, comparable, third-party evidence, which is precisely what a vendor's own site cannot be.
Learn moreReddit and community presence
Reddit is the most-cited domain in the analysed set, ahead of G2. Assistants, and ChatGPT in particular, lean on user-generated discussion for software recommendations, because it is where people say what a product is actually like to use.
Learn moreAnalyst and publisher placement
Five of the ten most-cited domains for B2B SaaS prompts are publishers or analysts: PCMag, TechRadar, Gartner, Forbes and TechCrunch. These are earned placements with long lead times and unusually long half-lives.
Learn moreYouTube and video
YouTube is in the ten most-cited domains for B2B SaaS prompts, and it is the one most SaaS marketing teams treat as a brand channel rather than a citation source.
Learn moreMeasurement, as a service rather than a bullet
Measurement is the part our positioning already promises and the part competitors find hardest to copy, so it is a service line rather than a line item. We have published the definitions behind every number we report, including the ways each metric is usually inflated, so you can hold us to them.
- A declared, frozen prompt set and competitor set, so the denominator cannot move
- Each prompt run repeatedly, because model outputs are non-deterministic
- Every platform reported separately, never blended into one score
- Follow-up turns measured, because visibility usually collapses by turn three
- Source concentration reported, so you can see which domains own your category