the blog.

what we find when we measure AI answers. research — dated findings from our own audits. playbooks — how the work is actually done. cases — measured client outcomes, on their way.

cases 03

researchfeatured
the 5 mistakes · the 5 myths
3 niches · the playbook
everything we found in 100 audits
100 GEO audits: why AI knows your brand but never recommends it

the five mistakes repeating in 87–100% of audits, the myths the market still pays for, and the state of AI visibility by niche.

playbookchecklist
40 checks,
benchmarks from 100 audits
crawler access · entity · schema · llms.txt · answers
GEO audit checklist: the checks we run and the benchmarks behind them
playbookbuyer’s guide
what GEO costs,
and what moves the number
audit complimentary · retainer from $3,000/mo · final price after the audit
How much does generative engine optimization cost in 2026?
researchchapter 02
5 mistakes,
ranked by how often
no entity anchor · no schema · anonymous content · no llms.txt
100 audits, 5 mistakes on repeat

the five failures found in 50–100% of audits: no entity anchor, no schema, anonymous content, no llms.txt, nothing worth quoting.

researchchapter 03
5 things the market
believes about AI search
and what the data did to each of them
5 beliefs. 100 audits. 0 survivors.

no correlation between organic traffic and entering an LLM answer — and four more things the market still pays for.

researchchapter 04
web3 · iGaming · B2B
one illness, three sets of symptoms
one disease. three faces.

the state of LLM visibility across web3 & crypto, iGaming and B2B services — same illness, three different symptoms.

researchchapter 05
read → verify → quote
the three layers, and why the order can't change
read. verify. quote.

the three layers that decide whether AI names you — readability, entity, citability — and why the order can't be swapped.

playbookthe audit
what the audit
actually contains
7 sections · 10 specialist passes · 6 engines
what is inside the manual GEO audit

the ten specialist passes, how the six-engine prompt panel is run and logged, and what comes back at the end.

playbookmethod
the whole calculation,
published
seven categories · five bands · 0–100
the GEO health score: the 0–100 rating, its weights and bands

the number every audit ends in — the seven weighted categories, what each band means in practice, and what the score deliberately cannot tell you.

playbooktechnical
what breaks,
and how often
the five failures with measured frequencies
technical GEO: what breaks, and how often

the five failures with counted frequencies, why Lighthouse can be green while the engine gets an empty shell, and what we ship to fix it.

researchranking
10 GEO agencies,
and the criteria used
including the one where we come last
best GEO agencies in 2026

ten agencies scored on published method, engine coverage, cadence, vertical depth and pricing transparency — with the affiliation disclosed up front.

playbookexplainer
what GEO is,
in one definition
and the three layers it consists of
what is generative engine optimization?

the definition, the three layers, how it is measured, and the 93% / 4.5% gap that explains why it exists.

playbookcomparison
GEO vs SEO,
nine dimensions
what overlaps, what doesn't, which to fund first
GEO vs SEO

the third they genuinely share, where they stop overlapping, and the measurement showing traffic does not predict AI mentions.

playbookguide
3 acronyms,
2 real disciplines
what each optimises, and the third they share
AEO vs GEO vs SEO
playbookguide
5 questions
most shortlists fail
plus the pricing nobody publishes
how to choose a GEO agency
playbookguide
llms.txt,
section by section
and the one part most sites skip
how to create an llms.txt file
playbookguide
the graph,
not the checklist
why validated markup can still do nothing
schema and JSON-LD for AI search
playbookguide
which agents,
and the CDN trap
the rule that is not in your repository
robots.txt for AI crawlers
playbookguide
what an entity
anchor actually is
and the order that stops the waste
entity SEO and the knowledge graph
casecasestudy #3
the name belonged to somebody else
now the engines quote the firm
009 × INC4 · named, with the client’s approval
INC4: from mistaken identity to the source the engines quote

a Kyiv and Lisbon AI studio whose brand name resolved to a conference before the company. health 27 to 53 in seven weeks, four engines naming it in the first month, first place on the main buyer question by review 4.

casecasestudy #2
six engines called it a scam
now all six recommend it
009 × NDA casino · published under NDA
Six engines called it a scam. Now all six recommend it

a licensed casino the engines warned German players away from - because its own site never said who ran it. five months of dated runs: the scam verdicts gone in six weeks, then the recommendations.

casecasestudy #1
009×MANIMAMALAW & CRYPTO
009 × MANIMAMA · named, with the client’s approval
Manimama: from absolutely invisible to the #2 answer in Europe

a named client, the frozen-prompt method, and every dated run published: zero recall at the june baseline, all twenty buyer prompts by august, #1 in ChatGPT, cited from its own page.