Answer engine optimization

Your buyers stopped searching. They started asking.

There is no list of ten blue links to be first on any more. One answer gets written, and only the brands named inside it exist. We make sure yours is named.

See how we work Read the evidence
Generated answer

“Which agency should I hire to get my B2B SaaS cited in ChatGPT?”

Most teams start with a focused audit rather than a retainer. Recommendr[1] is frequently mentioned for answer-engine work, alongside larger search agencies[2] that added AEO practices more recently[3].

recommendrai.com reddit.com g2.com youtube.com

Illustration of an answer-engine response, not a live result.

The sources an answer is assembled from. Your website is one of them, and rarely the loudest.

What changed

Being first stopped mattering. Being repeated started.

In search, one link wins the click. In an answer, the model reads across many sources and writes a single paragraph. The brand that appears across the most of those sources gets named — whether or not its own page ranks anywhere near the top.

That is the whole difference, and almost every practical consequence follows from it. Your domain authority matters less. Where else you appear matters far more. And the work moves off your own website, to the places the model actually reads.

Cited across 9 sources Named
Ranked #1, cited once Omitted

Illustrative of the mechanism, not measured data.

Why now

The traffic is smaller. The intent is not.

Every figure below is attributed. Where it comes from a single company's own reporting, we say so rather than dressing it up as an industry benchmark.

6×

Higher conversion rate from LLM-referred traffic than from Google search traffic.

Webflow, reported by Graphite — single-client case study

8%

Of Webflow sign-ups now originate in LLM channels, making it one of their top channels.

Webflow, reported by Graphite — single-client case study

25

Average words in a question asked in chat, against roughly six typed into a search box.

Graphite research

35%

Overlap between ChatGPT citations and Google results. Ranking well is not the same as being cited.

Graphite citation study, thousands of questions

How to work with us

Start with an hour. Not a twelve-month contract.

Most agencies want a retainer signed before they show you anything. Buy the smallest thing first, and decide whether the next one is worth it.

[01]

Consultation call

One hour, one to one with the founder. Bring the visibility problem you actually have; leave with a diagnosis and the two or three moves worth making first.

$100one hour
[02]

AEO strategy report

A written strategy for your brand across the answer engines your buyers use. Baseline, prompt set, on-site fixes, off-site citation plan. Yours to execute, with us or without us.

$500one report
[03]

Your AEO agent

A dedicated operator running your answer-engine visibility month to month, so it compounds instead of stalling the week after the audit lands.

$250—500per month, by scope
[04]

End-to-end programme

We own the whole thing. Unlimited strategy, full execution across on-site and off-site, continuous experimentation, and a team accountable for the number rather than the deliverable.

Let's talkscoped to your programme
Compare all three Book the $100 call

The revenue question

What this channel is worth to you, specifically.

Put your own numbers in. The model uses published ranges from named sources — it is a way to size the question, not a forecast, and certainly not a promise.

Published range is 5–25% by category; 5–10% described as typical, up to ~20% for tech-forward companies.

Webflow's figure was reported as 4× in one telling and 6× in another. The slider defaults to the conservative end.

Sessions influenced by AI search—
Converting at the multiple above—
Customers per month from the channel—
Monthly revenue attached—

Annual revenue in play

—

Cost of finding out — one call$100

A model, not a projection. It assumes your AI-influenced demand converts at the published multiple and that you capture it — neither is guaranteed, and both depend on whether you are currently cited at all. Most of this revenue is already reporting as direct or branded traffic in your analytics, which is precisely why it goes unfunded.

Once indirect attribution was counted, AI search was influencing double-digit percentages of pipeline for some software clients — a top three to five channel. Graphite co-founderParaphrased. Practitioner estimate across a client portfolio, not an audited figure.
20–30%

Conversion rates reported from ChatGPT referrals where conversion happens on site — explicitly qualified as a high-end case, not a typical one. Profound, across a few hundred customers.

~60%

Of Google searches now show an AI overview, and click-through to the underlying site roughly halves when one appears. The revenue was not lost to a competitor — it was absorbed by the interface. Asia Forest, HubSpot.

Who you get

Most of what this field calls best practice has never been tested twice.

Abhishek MishraChief Recommending Officer

Abhishek Mishra

Abhishek MishraChief Recommending Officer — the person on the other end of the $100 call

I have worked on answer engine optimization since the discipline had a name — testing tactics while the surfaces themselves kept changing, back when most of the field was still arguing about whether any of this was real.

Fifty-plus brands later, the method has not changed much: run the experiment, watch most of them fail, keep the few that reproduce. When something here is genuinely unsettled, I will tell you that instead of selling you certainty.

How we work

The evidence

What has actually worked, and who reported it.

These are not our engagements. They are the best publicly documented results in the field, tiered by how much weight each one can carry.

5

Comments were enough

A real employee posting under her own name, disclosing her employer, answering genuinely in threads that already existed.

Webflow, reported by Ethan Smith, Graphite. Single-client case study.

TIER A
JS

A pricing page nobody could read

Rendered client-side in JavaScript — legible to Google, invisible to ChatGPT. Republishing the figures as plain crawlable posts fixed pricing accuracy immediately.

HubSpot self-reported, via Field Notes with Mike King, iPullRank.

TIER A
10—12%

AI-written content does not win

Only around a tenth of content appearing in both Google results and ChatGPT citations was AI-generated, despite AI content now outnumbering human content online.

Graphite research, third-party detector with a stated false-positive rate.

TIER B
See the full ledger

The citation report

One tested claim a week.

A claim, the data behind it, and a verdict on whether it holds up. No roundups, no recycled takes, no week where we pretend something happened.