Coaching 02
Get Cited
By The
Machines.
Generative Engine Optimization. Your buyers are asking ChatGPT and AI Overviews the questions they used to type into Google — and the brands named in those answers aren't always the ones ranking first.
Engines Tracked
What GEO Actually Is
Generative Engine Optimization is the work of being named and cited inside AI-generated answers — ChatGPT, Claude, Perplexity, Gemini, and Google's AI Overviews — rather than merely ranking on a page of blue links.
The distinction matters because the mechanics differ in one crucial way. When someone asks an AI assistant “who are the best providers of X,” the model assembles its answer largely from third-party sources — listicles, review sites, comparison pages, forum threads. Your own website is frequently not among them.
Which means the instinct that works in SEO — publish a better page — often does nothing here. If the sources being cited don't mention you, another page on your own domain changes nothing. That reframes the work from publishing to placement, and it's the single most common thing teams get wrong when they start.
There's a second failure mode worth naming: factual drift. Models confidently state the wrong founding year, the wrong location, the wrong service list. That's usually an entity-definition problem — inconsistent schema, thin third-party profiles — and it's cheap to fix once you know it's happening.
What The Work Involves
A Real Prompt Set
25–30 prompts across five intent tiers — discovery, comparison, problem-first, buying criteria, branded — weighted toward the unbranded questions where visibility is actually won and where most brands score close to zero.
Multi-Engine Measurement
Run across ChatGPT, Claude, Perplexity and Gemini in clean logged-out sessions. Personalisation will happily show you your own brand and call it visibility.
Citation Analysis
The column that matters. Not whether you're mentioned, but what the answer cites — because that's where the fix lives, and it's frequently a competitor's content ranking your category.
Entity Definition
Schema, sameAs consistency, and third-party profiles, so the model knows who you are and states it correctly. Cheap to fix, and it's why some brands get described accurately and others don't.
Source Placement
Getting into the listicles and comparison pages the engines actually cite. This is digital PR with a target list derived from data, not a guess.
Tracking Over Time
Single-run results are noisy. Direction over months is the signal. Tooling covered on the Profound track.
How An Engagement Runs
Define The Buyer, Then The Prompts
Get this wrong and the whole audit is invalid. A white-label provider selling to agencies needs completely different prompts than a brand selling to consumers — and most prompt sets are written for the wrong buyer entirely.
Baseline Across Engines
Run the set clean and record presence, position, how you're described, competitors named, and every source cited. Expect a large gap between branded and unbranded performance; that gap is the finding.
Diagnose The Cause
Absent because the content doesn't exist? Because the entity is undefined? Or because the cited sources don't include you? Three different problems, three completely different fixes, and only the data distinguishes them.
Fix What You Control
Entity schema, extractable answer formats, and the definitional content models reach for first. Usually the fastest wins and almost always under-invested.
Earn The Placements
Target the sources that appear repeatedly across your prompt set. A domain cited on many prompts in your category has outsized influence on what every model says about you.
Re-Measure And Hand Over
Same prompts, same conditions, quarterly. Your team runs it. The method is published openly on GitHub — there's nothing proprietary to protect here.
Fit
Good fit
- checkYour category is one buyers research conversationally — software, professional services, considered purchases.
- checkYou rank well and suspect that's no longer converting the way it did.
- checkA competitor keeps getting named in AI answers and you don't know why.
- checkYou need a defensible baseline before committing budget to this.
Poor fit
- closeExpecting a guarantee of citation. Model outputs vary by session and by phrasing.
- closeHoping to skip the fundamentals. GEO does not rescue a site with no depth or authority.
- closeWanting a single number to report upward. Honest measurement here is directional, across many prompts.
- closePurely transactional local intent — if buyers search “near me” and call, this is a low priority for you.
Engagement Shapes
Baseline Audit
Fixed scope · one-time
A structured prompt set run clean across four engines, with presence, position, description accuracy, competitors named, and every source cited. Delivered as a written diagnosis, not a dashboard.
Audit & Placement Roadmap
Fixed scope · one-time
The baseline plus a ranked target list of the sources actually being cited in your category, and what it would take to appear in them.
Quarterly Tracking
Quarterly · ongoing
Same prompts, same conditions, every quarter. Direction over time is the real signal; a single run is noise.
Team Coaching
Monthly · ongoing
Your team learns to run the audit themselves. The methodology is published openly, so there's nothing to lock you into.
Pricing is set per engagement after a scoping call — scope, account count and cadence move it too much for a single published number to be honest. You will have a fixed figure before any work starts, and there are no open-ended hourly arrangements.
Common Questions
Is GEO actually different from SEO, or is it a rebrand?
Partly a rebrand, partly real. The content fundamentals are the same. What's genuinely different is that you're optimising to be quoted rather than clicked, the citation often comes from a third-party page rather than yours, and none of the standard tooling measures it.
Can you guarantee we'll be cited in ChatGPT?
No, and be sceptical of anyone who does. Model outputs vary by session, by phrasing, and by whether browsing is enabled. What can be measured is directional visibility across many prompts over time, which is what an audit establishes.
How is this measured?
A structured prompt set — typically 25 to 30 prompts across five intent tiers — run across ChatGPT, Claude, Perplexity and Gemini, in logged-out sessions, recording presence, position, description accuracy, and the sources cited. The last column matters most.
We rank first on Google. Isn't that enough?
Often not. I've audited a brand that appeared in five of nine results for its own name and zero times across thirty-two results on unbranded category questions. The site was fine. It simply wasn't in the third-party sources the models were citing.
Find out what
the machines say.
A baseline audit tells you where you stand across five engines — and, more usefully, why.
BOOK A GEO AUDIT