Proxy infrastructure×9.2 traffic
From 2,400 to 22,000 visits in two years, on links
IndustryProxy infrastructure
MarketsUS and EU
StartedMay 2024
Duration24 months
×9.2organic traffic
1,400referring domains
486AI mentions
Service · content for AI
AI content optimization
A model lifts two or three sentences, not the page. If there is no self-contained passage in the text, there is nothing to quote, however much you publish. So we start with an audit of what you have already published rather than with new articles, rework key pages to open with a direct answer and checkable numbers instead of adjectives, and your product team reviews every rewrite, because invented figures come back.
How much of this sounds familiar?tick what you recognise
Share of the library that cannot be quoted0%
Tick what you recognise – the estimate updates as you go.
A rough estimate from our audits, not a measurement. The exact figure comes after the review.
8–12pages reworked a month
6 weeksto the first citations
<30%of pages are quotable at the start
from $800a month
Why you are not quoted
Three reasons your text never gets lifted
Models quote passages, not pages
A language model lifts two or three sentences, not your whole article. That single fact reorders everything: a page can rank well and still never be cited, because there is no self-contained passage in it worth taking. We rewrite so that each section answers one question completely, without needing the paragraph above it.
Adjectives are unquotable, numbers are not
“Fast, secure and scalable” cannot be lifted into an answer, because it asserts nothing checkable. “Median latency of 40 ms across 12 regions” can. Most B2B content is written in the first register, which is why competitors with weaker sites get cited instead.
Rewriting beats publishing
You almost certainly already have the material. In our engagements, reworking an existing library produces citations faster and cheaper than adding new articles, because those pages are already indexed and already carry links. New content is the second step, not the first.
Formats
What gets lifted and what does not
Block format
Citation likelihood
Comparison table with named alternatives
high
Direct answer in the opening paragraph
high
Numbered steps with concrete values
medium
Definition with a cited source
medium
Feature list without numbers
low
Positioning paragraph before the answer
low
The last row is the commonest B2B mistake: the answer exists, but it sits under a paragraph about how remarkable you are.
How the work runs
From audit to new formats
Weeks 1–3Audit of the library
What is quotable and what is not
We go through the existing content and score every page on the things that decide citation: is there a direct answer, are there checkable numbers, is each section self-contained, are sources named. Most libraries score under thirty percent.
What you get
A scored inventory and a list ordered by which rewrites pay back first.
Months 1–3Rewriting
The pages that already rank
We start with pages that already hold positions, because a rewrite there changes citation without risking traffic. Direct answers moved to the top, comparison tables added, adjectives replaced with figures your team confirms.
What you get
Eight to twelve reworked pages a month, each reviewed by your product team before it ships.
Months 3–6New formats
What the library is missing
Comparison pages, glossary entries and structured explainers – the formats that get lifted most often and that almost no B2B library contains. Schema and llms.txt go in here too, though they matter less than the writing.
What you get
Citations start appearing in Perplexity first; AI Overviews follow by roughly two months.
A model lifts two or three sentences, not a page. That single fact explains why a site can rank well and never be quoted: there is no self-contained passage worth taking. Adjectives cannot be quoted, numbers can – and most B2B content is written entirely in adjectives.
Alexei Rozenplenter · Founder and CEO · About the author · LinkedIn
Case studies
What this did for tech companies' revenue
Read the case
ITAM / ITSM×4.4 traffic
From 1,500 to 6,600 visits a month in the US market
IndustryITAM / ITSM
MarketUnited States
StartedJanuary 2024
Duration24 months
×4.4organic traffic
71%non-branded
31%of AI answers
Edtech40,800 visits
Online exam platform: three locales that were blocking each other
IndustryEdtech
MarketsGlobal, IT, TR
StartedJune 2023
Duration13 months
40,800visits a month
3locales
×2.3clicks QoQ
Software development1,083 in the top 30
Developer directory: from nothing to 1,083 queries in the top 30
IndustryDeveloper directory
MarketUnited States
StartedNovember 2022
Duration12 months
1,083in the top 30
218in the top 10
−5,471left beyond-200
Figures are taken from the clients' own analytics for the periods stated.
How we do it
What actually changes in the text
Where the rewrite starts
The first move is scoring the library you already have, not writing new material. Every page is rated on four things: whether it answers a question directly, whether it carries checkable numbers, whether each section stands on its own without the one above it, and whether sources are named. Most libraries score under a third, which is good news, because reworking is cheaper and faster than producing.
We start with pages that already hold positions. A rewrite there changes citation while risking almost no traffic: the page is indexed, it has earned its links, and we are changing the form rather than the subject.
What changes in the text
The direct answer moves into the opening paragraph and the positioning moves down or disappears. Adjectives are replaced with values your team confirms: not "fast and reliable" but a specific latency under specific conditions. Comparison tables get added, because that is the format models lift most often and almost no B2B library contains it. Every section is rewritten so it can be read on its own.
Every figure is verified. An invented number survives until the first engineer on the client side reads it, and that costs more than having no number at all.
How this relates to ordinary search
These edits rarely hurt rankings and often help: a direct answer up top improves engagement, and self-contained sections raise the odds of a rich snippet. But that is not what we are aiming at. If you need growth in classic results, the work is site structure and links rather than the shape of the prose. That is the adjacent service, and we will say so if it suits you better.
AK
We'll score your library
Alex will take ten of your pages and show which passages a model could lift and which it could not. Free, and the document is yours.
Request a review
or call +375 33 331-90-76
Send us ten pages
The site address is enough, we'll pick them. We'll show what a model could take from your text and what it could not.