Publisher reporting, minus the theatre

Crawler logs, read as evidence.

Plumb turns crawler logs, analytics, and referral traces into editorial evidence. See which agents are reading, which stories are being reused, and where a policy or commission should land.

Schedule a demo Read methodology

window: 14 days
bots: 6 live
mode: publisher-owned

Fig. 1 — Crawler demand, 14-day trace

A living map of bot demand

measured

readingtoday · GPTBot

citation row34% CF · 1.4% Adobe

policy row5 active bot rules

counter4,010 req

top sourceGPTBot

volume, this day1,560

day-over-day+5.5%

Field note 01

AI consumption is already visible in your logs.

The point is not to celebrate the crawl. It is to turn the crawl into a line item your editorial, revenue, and policy teams can read in the same meeting.

Cloudflare34%

of origin requests are AI

edge logs / first-party evidence

Adobe Analytics1.4%

of sessions come from AI

analytics surface / lower than crawl volume

Plumb analysis24×

consumption-to-referral ratio

publisher-owned evidence trail

Workbench

A desk built around the data you already own.

Connect edge logs and analytics. Plumb shows crawler activity, referral gaps, and the evidence behind each recommendation before you change a policy or commission a story.

live feed12,847

origin requests per second

Origin requests / sec12,847+8.4%

Citation share (live)412.6%+12.1%

Active crawlers237+6

Bot demand · live request rows

14 days

Domain	Bot	Path	Requests	Note
crustwallet.com	GPTBot	/rates/crust	1,284	edge / allowed
deepdishrate.com	ClaudeBot	/compare	942	citations rising
forcheese.com	PerplexityBot	/best-rates	788	sampled
pizzapedia.com	Google-Extended	/guides	611	reference heavy
creditcrust.com	Applebot	/premium/*	402	policy flagged

Field note 02

A request path, not a pitch deck.

Plumb sits where the request becomes evidence and the evidence becomes a report a publisher can act on.

requestagent hit
the crawler asks for a page, a segment, or a feed
edgepolicy check
the request passes through the publisher’s control point
logfirst-party trace
Cloudflare, Fastly, GSC, GA4, or server logs keep the trail
Plumbreconciliation
telemetry, citation share, and policy are normalized together
reporteditorial brief
the output lands as a decision-ready brief, not a dashboard

Meter · verify · commission

Measure, explain, act.

Plumb gives teams a shared read on crawler demand, referral gaps, and the next pieces worth fixing.

Meter anything in one line.

Connect Cloudflare, Fastly, or server logs. Plumb cleans the stream, groups crawler activity, and keeps the raw trail close enough to audit.

Read the integration guide

Source	Detail	Share
Cloudflare	origin edge	34%
Fastly	cache misses	22%
GSC	search discovery	18%
Server logs	raw path hits	100%

Verify before you act.

Label every metric by evidence tier: measured, sampled, or inferred. Teams can see what is fact and what is a model.

See confidence tiers

Tier	Detail	Source
Measured	first-party trace	logs you own
Sampled	probe-backed	edge observations
Inferred	modeled relationship	Plumb analysis

Commission from evidence.

Every recommendation points back to the evidence. When an editor asks “why this?”, the answer is already attached.

Browse commission engine

BRIEF-04210

commission queue

Refresh$4K/yr

BRIEF-04211

commission queue

Retire$2.8K/yr

BRIEF-04212

commission queue

Update$6.2K/yr

BRIEF-04213

commission queue

New$12K/yr

Agent access

Plumb speaks MCP.

Every table, meter, and commission workflow in Plumb is also a tool call. Point Claude, Cursor, or your own agent at Plumb and ask the same questions your editorial desk does.

MCP

Fifteen tools, five resources, one token.

Mint a scoped token from the dashboard, drop it into your MCP client, and ask “what should I commission this week?” Plumb pulls bleeding queries, drafts the brief, and cites the evidence tier behind every claim.

{
  "mcpServers": {
    "plumb": {
      "command": "plumb-mcp",
      "env": { "PLUMB_MCP_TOKEN": "plumb_mcp_…" }
    }
  }
}

Open the MCP console

Name	Kind	Returns
list_bleeding_queries	Tool	AI Overviews eating clicks, ranked by priority
get_article_health	Tool	Citability + structural diagnosis for one URL
list_bot_crawlers	Tool	Crawlers hitting your content, policy state
get_visibility_panel	Tool	Share of voice by platform, week over week
create_commission_draft	Tool	Draft a commission from a query, brief attached
plumb://methodology	Resource	Measured / sampled / inferred, defined

+ 10 more tools, 4 more resources — see the full catalog in the console.

Prompt index

Questions editors actually ask.

The prompt set is a working index, not a slogan wall. Hover a row to see the reporting posture change when the evidence is already there.

01What changed in AI citations this week?on file
02Which topics lost referral share?on file
03Where are competitors replacing us?on file
04Which prompts trigger premium intent?on file
05What content should be refreshed now?on file
06Which crawlers need policy updates?on file
07What is MEASURED vs inferred?on file
08Which briefs deserve a commission?on file

Buyer fit

Who this is built for, row by row.

A capability matrix, not a comparison block. Each line ties a buyer type to the telemetry or workflow Plumb can actually support.

Buyer	Claim	Source
Built forCommerce	Article-level economics and affiliate revenue.	Revenue loop
Built forB2B media	Paid newsletter or subscription LTV.	Publisher model
Built forRevenue ops	Audience, SEO, product, and editorial planning.	Shared evidence
Built forInstrumented	Cloudflare, Fastly, GSC, GA4, or server logs.	First-party data
Not built forBrand sentiment	Generic share-of-voice dashboards.	No revenue loop
Not built forContent farms	Autonomous article generation workflows.	Wrong buyer
Not built forGeneral news	No article-level revenue loop to close.	No commission path
Not built forMarketplace-only	No first-party telemetry to reconcile.	Missing source data

Schedule a field test

Publisher insights on AI agents.

Connect your logs and analytics, choose the evidence tier you trust, then send a report your editorial, product, and revenue teams can actually use.

Schedule a demo Read methodology

Questions that usually arrive first.

What data do I need?

Cloudflare, Fastly, GA4, GSC, or server logs. The more first-party surface area, the stronger the measurement.

Does Plumb need a licensing deal?

No. Plumb verifies marketplace behavior with or without one.

What is measured versus inferred?

Measured comes from systems you own. Sampled comes from probes. Inferred comes from modeled relationships.

Is this for brand tracking?

No. It is for publisher revenue operations and editorial decisions.

Can teams export evidence?

Yes. Briefs, policies, and audit-grade evidence are designed to move outside the product.

Will it replace editorial judgment?

No. It gives teams a clearer instrument for making the calls they already own.