GEO · B2B SaaS & DTC ecommerce
AI already recommends someone
in your category. I make sure
it’s you.
When your buyer asks ChatGPT which tool to use — or which brand to buy from — it names three companies and cites its sources. I find out whether you’re one of them, then engineer the signals that change the answer.
Fill the form — Five fields, 60 seconds.
Measured across
- ChatGPT
- Claude
- Gemini
- Perplexity
- Google AI Overviews
The problem
Four ways you’re losing buyers right now.
01
You’re not in the answer.
Buyer asks for the best option in your category. Three competitors get named. You don’t.
02
Ranking #1 doesn’t save you.
AI builds answers, it doesn’t rank links. You can own page one of Google and be invisible here.
03
The models are wrong about you.
When you do get mentioned, the description is often outdated, or aimed at the wrong buyer.
04
Nobody sends you a report.
Unlike a ranking drop, this is silent. You find out when a deal you never knew about goes elsewhere.
Why this matters now
The shortlist is being built without you in the room.
Your buyer researches, compares and decides inside an AI answer, long before they ever land on your site. By the time they show up, the decision is mostly made.
- people use ChatGPT every week
- 900Mpeople use ChatGPT every weekOpenAI, February 2026
- of B2B buyers used generative AI in their purchase process
- 94%of B2B buyers used generative AI in their purchase processForrester Buyers’ Journey Survey 2026, ~18,000 buyers
- use it specifically to compare vendors against each other
- 55%use it specifically to compare vendors against each otherForrester, January 2026
- of the time, the vendor a buyer already favours before contacting sales goes on to win the deal
- 80%of the time, the vendor a buyer already favours before contacting sales goes on to win the deal6sense, 2025
Read those together: AI builds the shortlist, and the shortlist decides the deal. If you’re not named in the answer, you’re not in the running — and nothing in your analytics will tell you it happened.
Sources: OpenAI (Feb 2026)Forrester Buyers’ Journey Survey 2026Forrester (Jan 2026)6sense (2025)
What this is
Generative Engine Optimization is not SEO with a new name.
Getting your brand cited, recommended and correctly described when a buyer asks an AI a question in your category.
Google ranks pages. ChatGPT, Claude, Gemini and Perplexity assemble answers. The signals behind those two things barely overlap.
What decides an AI recommendation: whether a model can lift a clean answer out of your pages, whether your entity data is consistent everywhere it’s read, and which third-party sources the model pulls from in your category. A standard SEO audit measures none of it.
The difference
Two different games. Most agencies are only playing one.
If your current agency added “GEO” to the retainer without changing the methodology, here’s what they’re missing.
| Dimension | Traditional SEO | AI search visibility |
|---|---|---|
| Where you compete | Google results page | Inside the answer itself |
| What decides the winner | Backlinks, keywords, page authority | Citation frequency, entity consistency, source authority |
| What the content must do | Rank for a query | Survive passage-level extraction as a standalone answer |
| What you measure | Rankings, impressions, clicks | Share of voice per query cluster, across engines |
| How much you control | Your own pages | Your pages and the third-party sources models sample |
| How you displace a rival | Outrank them | Out-cite them |
| How stable it is | Ranks move gradually | Non-deterministic — same prompt, different answer, run to run |
| Buyer stage | Research | Shortlist — much closer to purchase |
| Does Google ranking help? | It’s the goal | It helps. It’s nowhere near enough. |
The method
The Citation Loop.
Four stages, running on a fixed cycle. Each pass compounds on the last.
01
Measure
A prompt matrix built from your category’s real buying language — informational, comparative and commercial intent — executed across five engines with repeated sampling per prompt, because these systems are non-deterministic and one run is noise. Every response logged, every cited URL captured.
02
Structure
Your pages rebuilt for passage-level extraction: heading hierarchy, self-contained answer blocks, claim density, internal consistency. Then the structured data layer — schema wired into one entity graph, with identity signals consistent across every source a model reads.
03
Source
Placement on the specific third-party pages your citation map proves are being pulled from in your category. Review platforms, category listicles, directories, community threads with real retrieval weight. Targeted by evidence, not by domain rating.
04
Re-measure
The identical prompt matrix, re-run on a fixed interval, scored against the baseline. Share of voice, citation count, and how the models describe you — all tracked as a delta, not a snapshot. Then the loop runs again.
Deliverables
What actually lands in your inbox.
Every engagement produces the same five artifacts. All of them are yours to keep.
01
Visibility Baseline
A prompt matrix built from your category’s actual buying language, spanning informational, comparative and commercial intent, executed across five engines with repeated sampling per prompt. Every response logged and scored: were you named, in what position, and in what framing.
- Prompt matrix
- Repeated sampling
- Position & framing scoring
02
Citation Source Map
Every URL the engines cited, deduplicated by domain, resolved to its owner and ranked by citation frequency across your query set. Split three ways: sources you control, sources you can earn, sources you can’t touch. That split is what makes the plan actionable instead of aspirational.
- Domain resolution
- Frequency ranking
- Ownership split
03
Retrieval Engineering
Your highest-intent pages rebuilt so a model can lift a clean, correct answer out of them — heading hierarchy, self-contained answer blocks, claim density, internal consistency. Plus the structured data layer: Organization, Product, FAQ and comparison schema wired into a single entity graph with consistent identity signals across every profile a model reads.
- Passage-level rewrites
- Entity graph & schema
- Comparison architecture
04
Off-Site Source Placement
Placement work aimed only at sources your citation map has already proven are being pulled from in your category — review platforms, category listicles, directories, community threads. Every target justified by evidence from your own data, not by a domain rating.
- Evidence-led targeting
- Review platforms
- Listicle & community
05
Measurement & Crawl Monitoring
The same prompt matrix re-run on a fixed interval so month-over-month deltas actually mean something. Plus server-log monitoring of AI crawler activity — GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended — so you can see retrieval demand, not just outcomes.
- Fixed-interval re-runs
- Share-of-voice delta
- Crawler log analysis
Built and run on tooling I wrote myself. No reseller dashboard, no white-labelled third-party report.
How it runs
From invisible to cited. Clear outputs at every stage.
Days 1–14
Baseline and source map
The full measurement pass and citation map. A document you can take to your team before another dollar is spent.
Days 15–30
Retrieval work shipped
Entity graph, schema layer, rewritten pages and comparison architecture — live, not recommended.
Days 30–60
Source placement
Placement executed against the targets your own citation data identified.
Day 30, then monthly
Re-measured
Identical prompt matrix, re-run and scored against baseline. One page: what moved, what didn’t, what’s next.
Honest filter
This works when the foundation is right. It doesn’t when it isn’t.
Not a fit if you’re
- Local services, or anything transactional with no research cycle
- Pre-product-market-fit and still working out who the buyer is
- Sitting on fewer than ten pages of real content — there’s nothing to extract yet
- Expecting to be the top recommendation in 30 days
- Looking for a visibility score as an end in itself
A strong fit if you’re
- B2B SaaS or DTC ecommerce in a considered-purchase category, where buyers compare before they commit
- Hearing “I found you through ChatGPT” but unable to reproduce it
- Already running SEO and watching AI Overviews eat the clicks
- In a category where competitors haven’t touched this yet
- Willing to actually ship the changes, not just receive a deck
Start here
Four fields. Then I go and look.
No call to book, no deck, no discovery process. You tell me your category, I go find out what the engines are saying, and you get it in writing.
01
You fill the form
Name, company URL, category, email. Sixty seconds.
02
I run your category
A prompt set built from your buyers’ real questions, executed across five engines by hand, with every cited source logged.
03
You get it in writing
What the engines said, who they named instead of you, and what it would take to change that.
Opens a Google Form. Results within 24 hours. No sequence, no follow-up unless you reply.
Pricing
One offer. One price. No tiers.
Month one covers the build. Every month after is $300. Commit longer, pay less across the board.
$270/month$300
after month one
Save $100
- Visibility Baseline
- Citation Source Map
- Retrieval Engineering
- Off-Site Source Placement
- Measurement & Crawl Monitoring
Why this is priced below the agencies: I’m building a track record, and the rate reflects that rather than the scope. The work above is what you get at any price — I’d rather have your case study than your budget. This is what it costs right now.
Larger scope quoted by email. I take on a small number of clients at a time.
Common questions
What people ask before starting.
How is this different from SEO?
SEO gets a link ranked. This gets your brand named inside the answer. The two overlap on content quality, but the deciding signals differ: whether a model can extract a clean passage from your page, whether your entity data is consistent everywhere it’s read, and which third-party sources get sampled in your category. A company can rank first on Google and be absent from every AI recommendation in its category.
Why is this so much cheaper than an agency?
Because I’m building a track record, not because the scope is smaller. You get the full measurement pass, the citation map, the on-site retrieval work and the off-site placement — the same artifacts an agency bills five figures for. The difference is that I’m one person with tooling I wrote myself and no overhead, and I want case studies more than I want margin right now. That won’t be true in six months.
Do we need to get on a call?
No. The whole engagement runs over email, including scoping and reporting. You send the form, I send findings, and we go from there in writing. If you’d rather talk it through, DM works too — but nothing here depends on finding an overlapping hour in two calendars.
What happens after I submit the form?
I build a prompt set from your category’s real buying questions, run it across five engines by hand, and email you what came back — including which competitors got named instead of you and which sources the engines cited to get there. If it looks fixable, the proposal is in the same email. If it doesn’t, I’ll tell you.
How long until it shows results?
Citation movement is typically measurable in 60 to 90 days. On-site retrieval work registers faster because retrieval reflects your live pages. Off-site is slower and depends on crawl timing. At day 30 you have the baseline and a full re-measurement — a concrete deliverable before deciding to continue.
Can you guarantee I’ll be recommended?
No, and be careful of anyone who does. These are closed systems that change retrieval behaviour without notice, and they’re non-deterministic — the same prompt can return a different answer twice in a row, which is exactly why the measurement uses repeated sampling. What I guarantee is measurement against the same prompt matrix every month, and that the work gets shipped rather than recommended.
Can’t my current SEO agency just do this?
Most have added GEO to the service list without changing the methodology. Ask them for live AI mention data from a current client across ChatGPT and Perplexity, and ask what sampling they use. If they can’t produce either, you have your answer.
Who am I actually working with?
Me, directly. No account manager, nothing subcontracted. That’s the trade — you get the person doing the work, and I take a small number of clients at a time.
What do you need from me?
CMS access or a developer who can ship changes, server log access if you want crawler monitoring, and a list of your real competitors. I write the pages and send them for approval before anything goes live.
Does this replace my SEO?
No — it runs on top of it. AI systems reference well-ranked content heavily, so a strong organic presence accelerates this rather than competing with it.