Analytics

See how your Channel gets found — people visiting your Library and AI agents querying your MCP endpoint — with sources at three depths, AI referrals badged, and the questions agents asked.

The Access Analytics page at podhood.com/studio/<slug>/analytics shows your Channel's discovery outcomes — evidence of the content being found — split by the two audiences Discovery serves: People (visits to your public Library) and AI agents (retrieval against your Channel's MCP endpoint). It complements Comments, which shows what the audience says once it has found you.

The numbers are outcomes of the surfaces PodHood publishes for the Channel: search visits land on the crawlable Library pages, AI-badged referrals are click-throughs from answers that cited them, and MCP calls run against the Channel's own endpoint. Get found & cited describes those surfaces; this page shows what they bring back. YouTube-sourced and RSS-sourced Channels get the same dashboard — for a YouTube channel it covers the half YouTube Studio can't see: discovery happening outside the platform.

Availability

Every Plan, including Free. There is no capability lock — the numbers are the product's proof of work. Connect a Channel and the page fills in on its own.
No Credits. Analytics never draws from the Owner's Credit pool. See Plans & credits.
Nothing to install. PodHood serves your Library and your MCP endpoint, so measurement ships built in — on your <slug>.podhood.com subdomain and on a custom domain alike.

Reading the page

Time ranges: 24h, 7d (default), 30d, and 90d. The 24-hour view buckets hourly; the others bucket daily.
Δ deltas: every headline number is compared against the completed period before it — on 7d, the current window (six complete days plus today so far) against the seven complete days before that. Until the newest bucket closes, deltas lean slightly conservative.
Refreshed hourly. The numbers are a deliberate snapshot, not a live feed; the header shows exactly how old the snapshot is, and a refresh lands within the hour.
The bucket in progress. The newest chart bucket — today, or the current hour on 24h — is still filling in; the trend chart fades it behind a dashed boundary so a partial bucket never reads as a drop.
Δ needs a sample. A delta chip only renders once the previous period is large enough for a percentage to mean something; small bases stay chipless instead of shouting ±triple digits.

People · Web analytics

The People half covers visits to your public Library, on every host you serve — subdomain and custom domain both.

Headline tiles — Unique visitors, Pageviews, and Avg. session duration, each with its Δ.
Top pages — the most-visited Library pages by pageviews, resolved to real episode titles so you can see which episodes pull visitors.
Sources — where visits came from, at three granularities:

Granularity	What it answers
Acquisition channels	The classified rollup — Direct, Organic Search, Referral, Paid.
Referring domains	The raw, browser-reported answer, domain by domain. No referrer reads as Direct; your own hosts never list.
UTM parameters	The self-declared answer — tags on links you shared, split by source, medium, campaign, term, and content.

AI source tag — a referrer or UTM row whose value matches a known AI answer engine (chatgpt.com, perplexity.ai, claude.ai, and more) wears an AI badge: visible evidence that an AI answer cited your content and someone clicked through. The badge annotates the row; it never changes a count.
Top countries — visitors by geography.
Devices — sessions by device type: Desktop, Mobile, Tablet.
Session quality — Bounce rate, Avg. session duration, Pages per session, and Sessions.

AI agents · MCP analytics

The Agents half covers AI agents querying your Channel's MCP endpoint.

Headline tiles — MCP tool calls and Agent connects (MCP handshakes), each with its Δ.
What agents asked — the search queries agents sent, verbatim, newest first. This is demand phrased in the asker's own words: questions your catalog is being asked before a visit ever happens.
Agent clients — which AI clients connect (Claude, ChatGPT, WorkBuddy, Codex, and others), identified from the MCP handshake. Agents that report an internal codename (for example WorkBuddy's per-connector names) are recognized and listed under their brand.
Tool health — calls, error rate, and p95 latency per tool, collapsed by default: it is service diagnostics rather than a creator metric.

If this half is empty, no agent has queried the Channel yet — share your Connect AI setup so agents can reach it. See Get found & cited.

UTM parameters are how a link declares where it came from. Add them to the links you control and the Sources card attributes those visits precisely:

https://<your-library>/episode?utm_source=newsletter&utm_medium=email&utm_campaign=ep142-launch

Pick a small, consistent vocabulary — utm_source for the place (newsletter, x, discord), utm_medium for the mechanism (email, social), utm_campaign for the push (ep142-launch) — and reuse it.
Untagged links still count; they are attributed by referrer instead.
Some AI answer engines tag their outbound links themselves — ChatGPT appends utm_source=chatgpt.com — so citation click-throughs surface here without any work on your part.

What's counted, and what isn't

Your team's Studio activity never counts toward audience numbers — People metrics cover the public Library only.
Self-referrals are excluded — navigation within your own Library is not a source.
Production only. Preview and development environments never pollute a Channel's numbers.
Aggregate, read-only, and rename-safe — history is anchored to the Channel itself, so renaming your slug never resets it.
Not downloads. Your podcast host keeps counting downloads and plays — that's consumption. Access Analytics measures the discovery step before it.

Was this page helpful?