woods · trywoods.ai/enterprise

For enterprise estates

Know what your estate is saying.

woods reads every public page, maps the estate, lists every claim it makes and shows where your content disagrees with itself. Each finding arrives with its evidence.

Get the Assessment

Public pages only. No sign-in, no credentials, nothing for IT to approve.

Six agents read the estate.

What each does today, and nothing it does not.

ScoutReads the public estate and maps every page, section and link.
WardenChecks structure and machine-readability. Groups faults by template.
RitaDates each page on evidence, names the evidence, and says “not known” when it cannot know.
SageExtracts every checkable claim in its exact words, with its page. Does not decide whether a claim holds.
CedarShows where two pages disagree, both versions side by side.
SamanthaScores each page’s voice against the estate’s own baseline.

Not yet measured: who looks after each page.

01ScoutEstate, Structure

See the whole estate in one picture.

#inventorytrywoods.ai/enterprise#inventory

The problem: Nobody can say how many pages exist.
Why: Tracking variants, redirects and old sections inflate every count.
What woods does: Scout maps every page, section and link, then states the distinct-page count and how it was deduplicated.
What you get: One inventory to plan from.

745

pages read and mapped, with every section and link.

293 pages no link on the site reaches.

grade A · case file 01 · a UK health brand · published 5 Oct 2026 · n = 1

02WardenStructure

Fix the template, not every page.

#templatestrywoods.ai/enterprise#templates

The problem: Faults look like hundreds of separate problems.
Why: One template repeats one fault on every page it renders.
What woods does: Warden groups orphans, dead ends and machine-readability faults by template.
What you get: A fix per template, not a fix per page.

No figure yet

No figure yet. the case file does not group faults by template, so this band shows no proof rather than borrow one.

03SageReports

Every claim, with the page it sits on.

#claimstrywoods.ai/enterprise#claims

The problem: Legal cannot list what the site claims.
Why: Claims sit in titles, descriptions and body copy, written by many people over years.
What woods does: Sage lists every checkable claim in its exact words, with its page and where in the page. Whether a claim holds stays your judgement.
What you get: A list legal can work from.

4,471

claims extracted in one sweep, each in its exact words with its page.

grade A · case file 01 · a UK health brand · published 5 Oct 2026 · n = 1

04Cedar and SamanthaReports

One story, everywhere it’s told.

#storytrywoods.ai/enterprise#story

The problem: The brand says different things on different pages.
Why: Positioning moved on. Older pages did not.
What woods does: Cedar sets the versions side by side, ranked by exposure. Samantha scores each page’s voice against a baseline drawn from the estate’s own pages. Which version is current is your call.
What you get: A ranked list of where the story splits.

153

contradictions across 4,471 claims, 24 of them high severity.

A customer count, two ways: One page says over 500,000 customers. Another says over 800,000. Both live.

grade A · case file 01 · a UK health brand · published 5 Oct 2026 · n = 1

05The crew togetherActions

A migration asks a question of every page. Put evidence under the answer.

#migrationtrywoods.ai/enterprise#migration

The problem: A migration forces Keep, Edit or Retire on every page, and most pages were never reviewed.
Why: Years of publishing with no retire step.
What woods does: Keep, Edit or Retire for each page, with one fix per template-level rule.
What you get: A worklist, not a reading list.

For a team in the middle of a migration →

488

open findings on 389 pages: 116 take minutes, 347 take hours, 25 need a decision.

grade A · case file 01 · a UK health brand · published 5 Oct 2026 · n = 1

06Rita and SageEstate

Know what your AI is reading.

#retrievaltrywoods.ai/enterprise#retrieval

Every AI system in your organisation is only as good as the knowledge it retrieves from.

The problem: An assistant or search feature answers from a page nobody trusts.
Why: Retrieval reads what is published, current or not.
What woods does: Each page is dated on evidence, with the method named and “not known” stated where it is.
What you get: A dated list of what your AI can read.

153

stale pages, and 188 too thin to answer a question.

grade A · case file 01 · a UK health brand · published 5 Oct 2026 · n = 1

07 Evidence, not opinion

Show the working.

#evidencetrywoods.ai/enterprise#evidence

Every finding is four parts: the observation, the reason, the evidence and its grade, and what woods could not see. One finding, opened up.

Showing the observation: what woods saw, stated as plainly as it can be. Choose another part to see how a finding is built.

Finding · Cedar · coherenceA customer count, two wayscase file 01 · grade A · n = 1

observed

500,000 or 800,000

One page says over 500,000 customers. Another says over 800,000. Both live.

An observation does not change between reads unless the pages change.

inferred

Two live pages give different figures for one fact, so a reader or an assistant cannot tell which is current.

This is inferred, not observed. It is allowed to change as more is read.

grade A

Both statements in their exact words, each with its page, found by Sage and compared by Cedar.

AOur own observation
BA source we have read and can name
CA secondary source
DNever published

This finding is grade A: pages woods read itself. Grades run A to C. D is never published.

From case file 01 · a UK health brand · published 5 Oct 2026 · n = 1. One estate, one sweep.

not seen

Which of the two is current. woods does not decide; the call is the reader’s.

Two statements with no wording in common would not be paired, so contradictions are a floor, not a ceiling.

Anything behind a sign-in, and anything outside the root address, was not read.

Solid is observed. Dashed is inferred or unseen. The inferences are allowed to change. The observations are not.

How a finding is built and graded →

08 Security and procurement

For the person who has to approve this.

#approvetrywoods.ai/enterprise#approve

Four plain statements. The security page gives each with its limit.

Read-only by design.No connector holds write scope, and that is not configurable.
Public pages only.The scope is a root address, and woods does not go outside it.
No page bodies stored at rest.Metadata, extracted claims and embeddings only.
No compliance claimed.woods supports governance and documentation. It does not make anyone compliant.

Read the security page →

09 The case file

One estate, read end to end.

#casetrywoods.ai/enterprise#case

case file 01 is one sweep of a UK health brand: 745 pages. It is shown with its limits, because a case file that shows only the good news is what woods exists to catch.

Read case file 01 →

153contradictions across 4,471 claims, 24 of them high severity
488open findings on 389 pages: 116 take minutes, 347 take hours, 25 need a decision
153stale pages, and 188 too thin to answer a question
293pages no link on the site reaches

case file 01 · a UK health brand · published 5 Oct 2026 · n = 1 · grade A

What this does not show

This is one sweep. The estate has not been swept twice, so there is no before and after here: this shows what a first read finds, not what fixing it changes.
The menu could not be measured. The crawl recorded no menu links on most pages, so no page was judged against it, and we say so instead of reporting a gap that is really ours.

Read your own estate.

The Assessment reads your public estate and returns the full report. It needs no credentials and no IT ticket.

Get the Assessment · £499

Public pages only. No sign-in, no credentials, nothing for IT to approve.

Anything else: the contact line is in the footer.

For the technical reader: the category is AI knowledge readiness.