Find out whether AI names you when buyers ask about your category, and what to fix so it does.
A buyer asks ChatGPT or Claude who to consider, and it writes back one answer with a few companies in it. No second page, no rank to climb. You’re in that answer or you’re not. You never see the buyers you lose this way.
51% of B2B software buyers now reach for an AI assistant before Google, up from 29% a year ago. 69% changed their pick based on what the assistant told them, and a third bought from a company they’d never heard of (G2, 2026). And the answer hardens. The longer you wait, the more the model leans on the companies it already names. Breaking in later costs more than getting named now.
AI doesn’t rank a list. It splits the buyer’s question into narrower ones, then builds one answer from whichever pages answer those best. On the broad term, you might sit behind the category leaders. On the specific question, where five companies compete instead of forty, the clearest answer wins, even from a smaller company. Win the specific question and you win the customer behind it.
You already paid for the content, the case studies, and the coverage that should make the model name you. It just isn’t getting through. The seven-day GEO audit finds why and hands you the fixes. Make them and that authority starts turning into pipeline. $3,500 fixed. Send a URL and a sentence on what you sell.
You either show up in it or three competitors get named in your place.
Each layer gets a 1-to-5 score, a plain finding, and a fix list ordered by impact and effort. The lower a layer sits, the harder it is to fix and the more it costs to leave it.
Roughly 68% of the enterprise B2B sites I’ve checked accidentally block at least one major AI crawler. The usual culprits are robots.txt, a Cloudflare or CDN rule, or how the site renders. If GPTBot or ClaudeBot can’t load your pages, nothing else here matters.
This is the structured data that tells a model what you are: Organization and Service schema, sameAs links, FAQ and Person markup. Half-filled, generic schema can read worse to AI than none, because it sends mixed signals about who you are. Most sites are running exactly that.
The mechanics here: answer-first structure, real heading hierarchy, FAQ blocks, timestamps, alt text. Most SEO work checks your pages for keywords. This one checks whether a model can pull a clean answer off them, which is where twenty years of information architecture pays off.
The signals that tell a model you’re real live off your own site: Wikipedia and Wikidata, Crunchbase, G2, Reddit, trade press, your founder’s footprint. Branded search and third-party mentions track AI citations more closely than backlinks. This is the slowest layer to fix, so catching the gaps early matters most.
I write 15 to 20 buyer-intent prompts for your ICP and run them live across the major assistants, screenshotting every answer. You see where you’re named, where you’re not, what the model says about you, and which competitors show up instead. This is the part that lands hardest, because it puts the answers buyers are getting in front of you.
When an assistant answers a question, it fires off a batch of longer, narrower sub-queries behind the scenes. Those show up in your Search Console, and most teams have no idea they’re there. Finding them tells you something concrete: whether AI is already pulling from your site, or skipping it.
Send a short email with your URL and two or three competitor URLs. First deliverable lands seven days later. Fit assessment and kickoff date come back within 24 hours.
Typical audit produces 20–40 findings. The list arrives already triaged so your team knows what to ship Monday vs. what to plan.
I record it personally, walking through the report in plain language. It’s the one thing a tool can’t hand you: a person telling you what to do and why.
The questions that come up on every GEO audit intake. Short answers; the full ones live in the report.
An SEO audit tells you where you rank on Google's ten blue links. A GEO audit tells you whether you show up at all when buyers ask an AI assistant about your category. They answer different questions.
Ask them which of the six layers they run. Most SEO teams are built for Google rank and haven't rebuilt their process for how AI retrieval works. What gets you cited there is extractability and entity authority. Position on a results page is a separate question. If they can name the six layers and show you their fan-out and citation work, use them. If you get a vague answer, that's the gap this fills.
It depends on your category. If you sell B2B software with a long, research-heavy buying cycle, AI citations are already shaping who makes the shortlist. If you sell something cheap and fast, the effect is smaller and further off. The audit gives you a straight read for where you actually sit.
The quick wins usually show up in 2–4 weeks. The slow ones, like building entity authority and earning third-party mentions, compound over 3–6 months. The audit tells you which fix is which.
If traffic isn't converting, CRO first. If you have decent conversion but suspect AI is recommending competitors instead of you, GEO first. Same price either way. If unsure, send a URL and get a straight answer back.
A URL and a sentence about what you sell and who you sell it to, a short kickoff call so I understand your buyer, and two or three competitor URLs. Read-only GSC access helps with Layer 6, the fan-out work. If it's slow or you can't get it, the audit still ships on time with five of the six layers. Layer 6 follows once access clears.
Yes, as a separate engagement, scoped after you've seen the findings. A lot of clients just take the fix list to their own developer or SEO team. Either way works.
Enterprise agency GEO retainers start at $3,000 a month. This is a senior practitioner running your category live across the major assistants and handing you a scored, prioritized fix list in seven days. Roughly the cost of one trade show badge.
A tool checks one of the six layers, the live test, and even that it gets wrong often enough to matter. Independent tests have caught trackers reporting a handful of mentions when the real count was over a hundred. The $2,000-a-month platforms just add more assistants to scan. None of them can tell you why the authority you paid for isn't reaching the answer, or what to fix first. A tool tells you that you're missing. It can't tell you why, and it can't tell you what to do about it.
Then you get a short report, a straight "don't spend on this right now," and the 30-day plan becomes a light monitoring cadence. We still do the debrief. You paid for certainty, and that's what you get.
Twenty years of B2B digital, including recent work for Moderna, J&J, Sapient, TATAA Biocenter, PlanHub, and Georgia Pacific. I'm NN/g trained with a Harvard GSD background. My most recent full CRO engagement delivered 10× organic traffic and tripled conversions.
You run marketing or growth at a B2B company doing $2M–$200M. You've already put money into traffic, SEO, paid, content, and the conversion side hasn't kept up. You want a diagnosis you can act on this quarter. A six-month engagement that hands you a deck in month three is the opposite of what you need.
Send a short email with your URL and two or three competitor URLs. First deliverable lands seven days later. Fit assessment and kickoff date come back within 24 hours.
