By Alex Price | July 2, 2025
Let’s cut straight to it: if you’re still chasing ‘position one’, there’s a chance you might be left behind.
GEO - Generative Engine Optimisation - doesn’t care about rankings. Because whilst AI tools like ChatGPT might make recommendations in a certain order, rankings in the traditional SEO sense don’t really exist any more.
B2B marketers need to shift mindset - from rankings to probabilities - as the rise of AI search engines takes place.
The old world of SEO: deterministic, trackable, predictable
Traditional SEO was a closed circuit.
You targeted a keyword.
You optimised a page.
Google showed ten blue links in a fixed order.
Your goal: get to position one and stay there.
The whole ecosystem - from Ahrefs to Semrush to every SEO agency on the planet - was built around rankings as the metric of visibility. You could track them. You could report on them. You could screenshot your client’s “#2 position” and pat yourself on the back.
That world has changed.
The new reality: GEO is probabilistic
Generative search doesn’t serve up fixed, repeatable, consistent results.
It generates answers in real time - often different answers for different people, or even different ones for the same person, depending on context, model memory, and randomness.
You might have noticed that the more you engage with ChatGPT, the more it remembers. That’s because it has a memory function.
Ask ChatGPT:
What are the top alternatives to Salesforce for a 500-person SaaS company?
Then ask Perplexity.
Then Claude.
Then Microsoft Copilot.
You’ll probably get different answers. Different vendors. Different justifications.
Sometimes links. Sometimes none. Sometimes citations (often very different citations depending on the LLM). Sometimes even hallucinations.
The results are not necessarily ranked, even if they are presented to you in a certain order.
They’re composed by models with constantly shifting training data and context sensitivity.
So what are you really optimising for?
You’re optimising for the probability of inclusion.
One prompt ≠ one output
In the world of SEO, there was broadly one search results page per keyword. OK, maybe it would change a bit depending on location or device (such as mobile) for certain search terms - but usually the ranking results would be consistent.
In GEO, the permutations are much wider:
- The exact prompt wording changes the outcome
- The user’s history/AI memory can influence the result
- The model’s “temperature” setting (i.e. randomness) affects what it says
- Forms of Retrieval Augmented Generation (RAG) pulls in live info - or doesn’t
- Plugins, web access, or third party integrations may further skew results
Every prompt is a dice roll.
Every answer is a composition.
Every mention of your brand is a probabilistic event, not a deterministic position.
Why AI visibility monitoring is key
In the old world, you checked your keyword rankings once a week or received an email alert from your rank tracker.
In GEO, things are dynamic in that models are retrained, updated, tuned regularly and at increasing pace.
And diverse, because there are so many AI tools currently being used (ChatGPT, Claude, Gemini, Perplexity, Slack copilots, custom GPTs etc etc).
And many of these tools now include memory.
That means if a buyer asked about CRM platforms three weeks ago, looked at some options and gave some feedback, ChatGPT might remember it and suggest different follow ups depending on that prior context. It might know about the buyer, their purchase history, current tools, preferences and more.
So if you’re not monitoring visibility constantly, across multiple tools, across multiple prompts, you are flying blind.
This is why platforms like Peec, our analytics partner at FirstMotion, are starting to emerge. And why we believe ongoing visibility auditing will become as fundamental as technical SEO audits used to be.
But you can still influence how AI tools decide to talk about you
Here’s the good news: while you can’t force an LLM to list your brand, in the same way you couldn’t force Google to, you can increase the probability that it does.
How?
- Third-Party Influence: Make sure you’re highly rated and well described on sites LLMs trust (G2, Capterra, Reddit, LinkedIn, Gartner, etc.)
- Structured, Citable Content: Use clear headers, concise answers, and semantic structure that models can extract easily
- Deep Audience Intelligence: It’s fundamental to understand your buyers, their pain points, drivers, goals, buying triggers etc - otherwise effective prompt mining isn’t possible
- Prompt Based Content Planning: Build content for prompts, not just keywords. Think: “How do I evaluate XYZ tools?” or “What should I ask a vendor in a demo?”
- Frameworks like PromptPath™: Map prompts across your buyer journey and shape visibility at every stage
- Ongoing Visibility Audits: GEO is not a one and done exercise - it’s a system you need to operate continuously
If you’re looking for a generative engine or AI search agency to help you understand how your B2B software brand is currently visible alongside competitors in tools like ChatGPT currently, get in touch.
Author Bio
Alex Price
I dropped out of university to turn my part time freelance business that I started as a teenager in my bedroom into an award winning digital agency. I grew the business organically, with no debt or investment, from just me as a 20 year old sole founder to a team of ~35 people and multi-million £ annual revenues and ~25% net profit margins - winning clients like Amazon. I also founded FINITE, a B2B marketing media brand and global membership community for software CMOs. In April 2022, aged 29, I sold the business to a US headquartered strategic buyer, realising the value of lots of hard work and achieving a life changing outcome.