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 · GEO and AI search optimization
AI Search Optimization
Generative engine optimization, answer engine optimization, AI SEO: whichever name you use, the job is getting ChatGPT, Perplexity, Gemini and AI Overviews to name your product. We start with a baseline: two hundred queries across four platforms, re-run every quarter, showing how often answers mention you today. Then we raise that share with content a model can quote and the external sources it actually cites. It rests on the same foundations as classic SEO – a page a crawler cannot read is one a model cannot quote. Included in our $1,500 SEO and AI search plan, or $800 on its own if your classic search already performs well.
We reply within half an hour. Any longer and we're at lunch.
Q“Which log management service should a mid-size SaaS company choose?”
For a mid-size team there are usually three candidates. Logstack wins on price below a terabyte a day, Traceloop offers deeper request tracing, and Beacon is simpler to adopt if you already run Kubernetes. The choice comes down to whether storage cost or detail matters more to you.
For a mid-size team the names that come up most often are Your product, Logstack and Traceloop. The first is singled out for transparent pricing with no retention surcharge and for taking around an hour to connect – both the documentation and independent comparisons say so. Logstack is still cheaper at very high volumes.
A real client, competitor names changed. Left is February, right is eight months later.
200queries in the baseline
4platforms
34%of answers name the client after 8 months
4–6 mo.payback
Why this is different
Three ways it differs from a normal SERP
There are no ten slots
In search, position ten still earns clicks. In an AI answer three or four products get named, and hardly anyone reads past the third. This isn't “another channel”, it's a channel with far harsher selection.
They cite structure, not polish
Models pull passages from pages with verifiable facts, numbers, sources and direct statements. Vague marketing copy gives them nothing to quote – there's no sentence you can lift into an answer.
The people asking are ready to buy
“Which tool should I pick” is a question from somebody with a problem and a budget, not a browser. That's why these visits weigh more in revenue than in traffic. It shows up most clearly in B2B and SaaS.
Price
What's in the fee and what isn't
Included in the fee
40 team hours a month.
- 200-query baselinequarterly
- Pages reworked for citation8 / mo
- Schema and llms.txt–
- Mention monitoringweekly
- Work on external roundups–
- Reportingevery 2 weeks
Separately, at cost
With the supplier invoice attached. These are averages across our projects. Your budget depends on how competitive the niche is and on what competitors spend in it; we size it in the link gap analysis before you commit.
- Placements in independent roundupsat the publisher's rate
- Developer hours$40 / h
The baseline can be bought on its own for $500, if you'd rather see the number before deciding.
How the year runs
From baseline to share of answers
Month 1Baseline
200 queries from your niche
We run two hundred questions your buyers actually ask through ChatGPT, Perplexity, Google AI Overviews and Gemini. We measure what share of answers name you, and who gets named instead.
What you see
A starting percentage and a named list of the competitors cited more often than you.
Months 2–4Rewriting for citation
Facts instead of vagueness
We rework key pages so a ready-made statement can be lifted from them: concrete numbers, comparison tables, cited sources, direct answers in the opening paragraph. Plus schema and llms.txt.
What you see
First mentions appear in Perplexity, which refreshes faster than the others.
Months 5–8Building presence
External sources
Models don't only pull from your own site. We work on independent comparisons, industry roundups and the places citations come from. In parallel we track how the wording about you changes.
What you see
The share of answers naming you, in the same report as every other metric.
Methodology
How we measure generative engine optimization
Brand mention rate
Share of answers in the sample that name you at all, with or without a link.
Citation rate
Share of answers that link to one of your pages as a source.
Competitor share of voice
How the mentions split between you and the vendors you are compared against.
Most cited pages
Which of your pages the models actually quote, which is rarely the ones you expect.
Commercial prompt coverage
Visibility on the prompts a buyer types when choosing a vendor, not on informational ones.
The same two hundred queries are re-run every quarter across the same platforms. Changing the sample would show a change in method rather than a change in result.
FAQ
What people ask about GEO, AEO and AI SEO
Measurable, which is why we start with a baseline. Two hundred queries, four platforms, a percentage on record: how many answers name you today. Six months later we run the same queries again. If the number hasn't moved, that's as visible as a traffic drop.
All four platforms we work with query the live web at answer time rather than relying only on weights. A fresh page can be cited within weeks. Perplexity picks things up fastest; Google AI Overviews is slower but sends more traffic.
They overlap: both need an indexable page with something useful on it. Then they diverge. Search rewards the page that answers the query best overall. An AI answer rewards the paragraph a statement can be lifted from cleanly. Those are different demands on the writing, and they often conflict.
It won't hurt and it's cheap, but don't overrate it: there's no settled standard yet and most platforms ignore it. We add it because it costs almost nothing, not because it solves the problem. What solves the problem is content structure and external mentions.
From $800 a month as a standalone service. It is already included in the $1,500 SEO and AI search plan, so buy it separately only if your classic search is already performing well. Faster than classic SEO, because “which tool should I pick” brings people closer to a decision. The volume is still smaller though – this supplements search rather than replacing it.
Both, in the sense people use the words. GEO, AEO, AI SEO and LLM SEO describe the same work: measuring how often AI assistants name you and raising that share. We call it AI search optimization and run it as part of our SEO plan or on its own. The differences between the terms are explained in GEO vs AEO vs AI SEO.
We can get you named more often, and we measure it on a fixed set of two hundred questions, but nobody can guarantee a position: answers change daily and differ by phrasing. Most of the result comes from being present on the third-party pages ChatGPT reads. The steps are in How to rank in ChatGPT.
How we do it
Answer engine optimization: how citation is measured and moved
GEO, AEO, AI SEO, LLM SEO: one job
Agencies sell this under four names. Generative engine optimization (GEO) comes from research on AI-generated answers; answer engine optimization (AEO) is the older term from featured snippets and voice search; AI SEO and LLM SEO are the marketing versions. The work behind them is the same, so we compare proposals by scope and measurement rather than by label. The longer version is in GEO vs AEO vs AI SEO.
What decides the result is mostly off your site. In our study of 152 sources that assistants cited when naming ITSM and proxy vendors, none sat on the vendors’ own websites; most were “best X” rankings published by someone else. That is why the plan includes work on external roundups, not only on your pages.
How citation is measured
The baseline runs on a sample of two hundred queries your buyers actually ask, not on generic industry phrasing. Each goes through four platforms, and we record more than whether you were mentioned: we record where in the answer you appear. Being named first, third or listed after the main recommendation are different outcomes, and almost nobody reads past the third name.
The same queries are re-run every quarter. That matters: changing the sample between runs destroys comparability and produces a chart that shows a change in method rather than a change in result.
What models take from a text
A model lifts a paragraph, not a page, so structure decides the outcome rather than length. Comparison tables with named alternatives work. So does a direct answer in the opening paragraph, concrete figures and references to checkable sources. Feature lists without numbers do not, and neither does a positioning paragraph sitting above the answer, because there is nothing in it to quote.
On llms.txt: we add it because it costs almost nothing, but we do not treat it as decisive. There is no settled standard and most platforms ignore the file. Content structure and external mentions decide this, not a file in the site root.
What to expect on timing
Perplexity picks up changes fastest, with first mentions usually appearing after six to eight weeks. AI Overviews trails by roughly two months but sends noticeably more traffic. ChatGPT is the least predictable of the three, because external roundups and directories carry weight there, and those move slower than your own site.
One honest caveat: the volume of this traffic is still smaller than search. It supplements organic rather than replacing it, and anyone selling you the opposite should be asked for numbers.
Being named in an AI answer is a different job from ranking, and the honest part is that the traffic is still smaller than search. We measure the same two hundred queries every quarter rather than changing the sample – otherwise the chart shows a change in method, not a change in result.
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.
SK
Alex will walk you through the numbers
Head of Inbound Sales. He'll ask about your deal size and margin, then tell you whether this service is worth it – including when it isn't.
Request a forecast
or call +375 33 331-90-76
Find out whether you're named today
We'll run two hundred queries from your niche and show you the percentage plus who gets cited instead. The first baseline is free.