A page of links became a paragraph.
Search handed the buyer ten options and let them choose. An answer engine hands them one paragraph naming two or three vendors. Everyone outside that paragraph is not ranked lower. They are absent.
Loading
About MoatWorks
Mumbai, India
AI visibility for B2B SaaS
We don’t optimize your website.We get you named on everyone else’s.
WE ARE MOATWORKS
Ask an answer engine who solves this and it names a competitor’s review before it names you. We work on the pages models actually trust, measure whether that changes, and say so when it doesn’t. That is the whole company.
Answer engineSchematic
›Which agencies handle AI visibility for B2B SaaS?
The agencies most often recommended for this are , and .(Three vendor names, shown redacted. This panel is a schematic.)
A schematic, not a screenshot. Those three names are there because the pages the engine read already named them. Winning that sentence means working on those pages, which is the part almost nobody is doing.
Measured againstChatGPTClaudePerplexityMicrosoft CopilotGemini & AI Overviews
Answer engineSchematic
›Which agencies handle AI visibility for B2B SaaS?
The agencies most often recommended for this are , and .(Three vendor names, shown redacted. This panel is a schematic.)
A schematic, not a screenshot. Those three names are there because the pages the engine read already named them. Winning that sentence means working on those pages, which is the part almost nobody is doing.
The premise
Why this company exists.
Search gave you
ten chances.
An answer gives
you one.
The paragraph is built from other people's websites.
Models reach for the sources they already trust: roundups, review platforms, documentation, threads. Your own site is one input among many, and rarely the deciding one. Optimising it harder does not fix a problem that lives somewhere else.
Almost nobody has measured where they stand.
Ranking is measured to three decimal places. Being named in an answer is usually measured by someone screenshotting one favourable reply. That gap is the entire reason this company exists.
Retrieval
Where the answer is assembled.
Five engines.
Five different
reflexes.
A brand can be named in every ChatGPT answer and absent from Perplexity, because the two are not reading the same shelf. Every audit here runs the same fixed prompt set against all five and scores them separately.
ChatGPT
Bing's index, roundups, review platforms, Yelp reviews, forum threads
Claude
Brave's index, documentation, primary sources, technical writing
Perplexity
Its own crawl, live search results, listicles, news
Microsoft Copilot
Bing's index, LinkedIn, Windows and Office surfaces
Gemini & AI Overviews
Ranking pages, structured data, entity records
Schematic — the kinds of source each engine leans on, not a measurement
Position
Contrarian, and checkable.
Nothing here
is a secret.
All of it
is checkable.
Every claim on this site can be checked by someone who is not us.
A flat month gets reported as flat. Nobody here is guaranteed a citation.
Three things the category sells that we will not
Named, not gestured at
The termSold asWhat is true
llms.txt
Shipped, never soldSold asPositioned as the file that gets you into AI answers.
What is trueThere is no public evidence that any major engine reads it as a ranking input. It costs us about an hour, so it ships in week one of every retainer and then we stop talking about it.
Schema-for-AI
Included, priced as plumbingSold asSold as a monthly line item with a strategy price attached.
What is trueStructured data genuinely helps an engine parse a page. It is the floor of the work, not the lever, and it does not need re-selling every thirty days once it is in place.
Content chunking
Not a serviceSold asSold as a retrieval technique with a diagram behind it.
What is trueWe have found no published evidence that reformatting your paragraphs changes what a model retrieves. Until someone produces it, charging for it is charging for a belief.
Client story
How the work holds in practice.
Case studyBaselineHerd Security
The best-regarded brand almost no one is shown.
When an engine names Herd it does so in the most purchase-ready terms in the category, and that almost never happens. The cause is not the product. It is visibility and identity.
Problem
Herd is coherent on its own domain and incoherent everywhere else. High-authority sources describe the same company four different ways: a security awareness platform, an API security tool, an endpoint protection vendor, and a voice fraud service. Six unrelated products share the name.
Solution
Technical access is the one column that is not the problem, and it scores 3.5. The budget belongs in structured data and off-domain presence, not in rebuilding a site that already retrieves.
Share of answers naming the brand11 of 13
- KnowBe446.7%
- Hoxhunt34.2%
- Proofpoint23.6%
- Adaptive Security21.1%
- SoSafe15.1%
- Huntress14.6%
- Usecure12.1%
- NINJIO10.6%
- Cofense7%
- Mimecast5%
- Herd Security3.5%
The baseline
13 to 27 Jul 2026 · 14 days3.5%
Average citation rate across 4 engines
1.6/ 5
Defensibility depth, weighted
< 2%
Share of voice, priority clusters
0/ 10
Priority query clusters won
How it was measuredNo testimonial
There is no quote here because there is no quote on record. What there is instead is the instrument: a fixed prompt set, a fixed competitor set, and the same run protocol every time, locked before the first measurement so the next re-test can only agree or disagree with this one.
- Sector
- Security awareness training
- Market
- United States
- Engines
- 4 measured · Claude scheduled
- Prompts
- 50 across 10 clusters
How we work
All four are in the contract.
Four terms.
All four are
in the contract.
Terms of engagementEvery tier
- Scope
Scope is locked before anything moves.
Fixed at the end of week two, against what the audit found. No change orders after that, and the number you signed is the number you pay.
- People
The people on the call do the work.
Placements and community participation happen under real names, by people who already have standing in those places. Nothing is offshored to a content farm.
- Guarantees
Nothing is guaranteed that cannot be.
Nobody outside the labs controls what an engine says. What is guaranteed is the work in the scope, a measured baseline, and a report honest enough to show a bad month.
- Ownership
You keep everything, including if you leave.
The assets, the entity work, the placements and the raw data are yours. There is no auto-renewal and no clause that makes leaving expensive.
The same four on every engagement, at every tier. Nothing here is negotiated away for a bigger retainer.
The studio
No account layer.
Three people.
The ones on the call
do the work.
Maitrey Gholap
Co-founder & CEO
Pranav Darekar
Co-founder & CFO
Raif Bashir
Chief Content Officer
WE WALK
AWAY
IF
Three reasons we refuse business.
01
Anyone who wants a guarantee.
The engines re-rank without notice. An agency promising you a citation is either guessing or lying, and we would rather lose the deal than join them.
02
Anyone buying volume.
If the brief is thirty blog posts a month, there are cheaper shops and they will do it better. This is a small number of assets built against prompts you are measurably losing.
03
Anyone who needs it this quarter.
Citations have to be published, crawled, then repeated by a model before they show up in an answer. That is a six to ten week arc on a good run, which is why nothing here is sold by the month.
The ask
A scoping call, not a pitch.
Find out what
the engines
say about you.
The audit is $800, takes seven business days, and comes off your first month if you go further. The first call is a scoping call, not a pitch.
Start with the measurement. Decide about the retainer afterwards.