How We Work

The Evidence Engine.
Build. Measure. Report.

The Evidence Engine is the system behind every Legency retainer: one accountable process that builds your site, measures what it earns, and reports it monthly in evidence a marketing leader can take upward.

See what an AI visibility read includes
The Evidence Engine

Three Stages,
Running on a Loop

Building the site is stage one. The Evidence Engine keeps going: it measures what the build earned, then reports it in a form your leadership can act on. Each report sets the next month of work.

01

Build

Your team ships without waiting on a developer

Requests land in a shared Slack channel and go out the same day where scope allows. Everything is built as reusable components, so volume gets cheaper as you go: one rollout put 205 new answers live across a six-site estate, each one tied to an article already published there.

02

Measure

You find out which channel actually earns the demos

On one account organic search overtook paid as the top source of demo requests: 39 people against 19 in a 28-day window. We track rankings, conversion on the pages that matter, and whether ChatGPT, Perplexity, Gemini and Claude cite you.

03

Report

A report you can forward without rewriting it

One a month, same format every time, numbers first. Our longest-running account has had 18 months of them, so month 18 compares to month one in ten seconds. Where a number moved the wrong way, it says so and says what changes.

Then it loops. This month's evidence writes next month's build plan.

What this looks like in numbers

0 to 475pages made machine-readable for AI engines, in two months
385pages given the search snippet they were missing
-83%crawler time lost to junk URLs, 2,349 down to 410
3 of 4AI engines that name a client for its speciality

Measured on live client engagements. Each figure stays anonymous until that client clears being named.

What We Measure With

The AI Readability Score

You get one number, 0 to 100, for whether ChatGPT, Perplexity, Gemini and Claude can actually use your pages when a buyer asks a question in your category. It is our own 100-point audit, and it is re-run line by line so the movement is measured rather than asserted.

A low score explains why a strong site goes unnamed when an assistant answers. The read finds the causes, puts them in the order worth fixing, and the number moves as those fixes land.

What a read includes

  • Every page, checkedHow many of your pages parse cleanly for a machine, and how many are invisible.
  • Snippets and metadataThe pages going out to search and AI engines with nothing to say about themselves.
  • Crawl wasteHow much crawler time is spent on junk URLs instead of the pages that earn citations.
  • Schema coverageWhether an engine can tell what each page is: a service, an article, a case study, a question.
  • A citation baselineWhere you are named today across ChatGPT, Perplexity, Gemini and Claude, and who is named instead.
  • A ranked fix listThe changes that move the score most, in order, with the effort each one takes.
A recent client moved from 30 to 59 in two months.
3059
0100
Procurement Readiness

Your Legal, Security and
Finance Teams, Answered

Reviews are where new agency relationships stall. We arrive with the paperwork already done, so yours takes days instead of quarters.

Security posture

Least-privilege access, credentials in a managed vault, and a named contact for your security team. Nothing reaches a production site without your sign-off - changes are staged first, every time.

Enterprise due diligence, passed

A 36-question enterprise cyber due diligence questionnaire, run by a client security team, answered and accepted. Your reviewers get straight answers instead of a sales deck.

MSA and SOW paper

A master services agreement and statement of work ready to sign, covering IP, subcontracting and confidentiality. We sign client paper too, and we have been through that review.

PO and AP billing

We bill into enterprise accounts payable with a purchase order number on every invoice. Your finance team gets what it needs the first time.

Webflow Premium Partner - Enterprise

The credential behind every build: the official Enterprise tier of the Webflow partner program.

The Engagement

How Working
Together Works

  • A retainer, not a project

    Monthly scope, agreed up front. Anything beyond it is quoted before work starts, so no invoice ever surprises you.

  • Slack-native, inside a written process

    Requests land in a shared channel and move through documented steps. Because the process is written down, the pace holds even when people are away.

  • Evidence on a fixed cadence

    One report a month, same shape every time. You can compare month four to month one in ten seconds.

  • Yours from day one

    Everything built under the retainer belongs to you immediately. Notice is 60 days after month three, and the work keeps running if you ever leave.

Your first month, week by week

  1. Week 1

    Access and baseline

    The Slack channel opens and access is granted under least-privilege rules. We take your AI Readability Score baseline and your analytics baseline, so month one has a start line.

  2. Week 2

    First builds ship

    The highest-value fixes from the baseline go into staging. They go live on your approval, not ours.

  3. Week 3

    Measurement wired

    Tracking is checked end to end so the numbers can be trusted. The citation panel is pointed at your domain across four AI engines.

  4. Week 4

    First report

    What moved, what it means, and the plan for next month. Written to be forwarded upward as it is.

By day 30 you are holding evidence you can forward to your leadership.