Content Intelligence

Content Intelligence looks at a page you own and returns a deterministic readiness score, the evidence behind it, a plain-language brief, and — if you opt in — an AI-generated draft. One analysis draws one content_analyses credit from your monthly allowance or your credit pack.

What it does

• Fetches the page through the Firecrawl provider you own the crawl for.
• Extracts headings, links, structured data, and a small set of on-page signals. No third-party HTML is stored raw.
• Scores the page against the same rule engine that powers the audit report, so the numbers stay comparable.
• Returns evidence for every finding: the rule id, where it fired, the citation excerpt, and a confidence label.
• If you opt in, calls the AI provider order from your operator config to produce a brief and (optionally) a draft. The provider selection is server-side; you never see raw provider payloads.

What it does NOT do

• It never publishes anything. Drafts are yours to accept, edit, or dismiss.
• It never crawls arbitrary URLs. The URL must belong to a site you own or a competitor you have added on the Agency tier.
• It never returns the raw provider payload, prompt, or completion. The public API response is the same normalized shape the app renders.

Starting an analysis

You can start an analysis from:

• Sites → any site → Content — pick a page URL from your inventory.
• Report → Fix now / Watch — the recommendation row opens the workflow pre-filled.
• Keyword research → tracked keyword — analyzes the URL that ranks for the keyword right now.
• Google Search Console → any query — analyzes the URL Google surfaces for that query.
• Competitors → Agency competitor page — Agency tier only.

Each entry point deep-links into the same workflow. Every one of them reserves capacity before enqueueing (see Plans, limits & credits); the flow refuses loudly if you are over your cap.

What one credit includes

One content_analyses credit covers:

• one page fetch through Firecrawl,
• the deterministic score and evidence,
• the plain-language brief,
• and — when opt-in AI is enabled — one draft.

Regeneration draws another credit. Retrying the same URL within the seven-day snapshot window reuses the cached page but still costs one credit for the new score and brief.

Score, evidence, confidence, citations, brief, draft

• Score. Deterministic 0–100. Same rule engine as the audit report.
• Evidence. Each finding names the rule, points at the exact HTML location it fired on, and shows a short excerpt.
• Confidence. high / medium / low. Confidence is a function of page structure quality — it is not vendor spin.
• Citations. Every AI claim cites the URL and excerpt it came from. Uncited claims are not shown.
• Brief. A short, plain-language plan you can share with a writer.
• Draft. Opt-in. The AI provider order in operator config decides who writes it. You accept, edit, or dismiss it — it is never auto-published.

Opt-in AI processing and retention

The AI brief and draft are opt-in. When you opt in, we send only the excerpts needed to produce the brief. Excerpts are retained for seven days so we can regenerate on request without a second fetch, then purged. Your export and account-deletion controls in Account security still apply and complete inside the documented grace window.

Partial and refunded results

Sometimes Firecrawl or the AI provider degrades. The analysis surfaces what completed successfully with a partial flag; you can regenerate the missing sections at no extra credit. If the run failed hard before any useful signal, the credit is refunded automatically and the analysis is marked refunded.

Recommendations and 28-day correlations

The brief lists recommendations. You can:

• Accept — moves the recommendation into your outcome loop.
• Dismiss — records the dismissal reason for the correlation window.
• Apply — marks it applied; the outcome loop measures the 28-day before/after correlation on rank, clicks, and impressions.

The correlation is descriptive, not causal — it is the same window we use in the audit report, and the number tells you whether the change moved.

Pro inventory and cannibalization

On Pro and Agency, the workspace opens a page inventory that groups URLs by topic and flags cannibalization pairs (two of your own URLs competing on the same intent). Inventory scans do not draw analysis credits — they run on the shared vendor cache.

Agency competitor portfolio and monitoring

Agency adds a competitor portfolio: you pick competitor URLs, run analyses against them, and — with Content Monitoring on — get notified when a tracked competitor page changes materially. Monitoring uses the Firecrawl change signature; a dismissed signature does not re-fire.

Managing allowances and the 20-analysis pack

Your monthly content_analyses allowance depends on your tier — see Plans, limits & credits. If you run out mid-month, the 20-analysis credit pack adds twenty single-use credits that never expire and stack on top of the monthly quota.

Troubleshooting

• Access. The URL must belong to a site you own (or an Agency competitor). Otherwise the flow returns "not found" — never "forbidden" — so it does not leak whether the URL exists.
• robots.txt. If the page blocks crawlers, the analysis returns partial and cites the block; we never bypass robots.
• Stale GSC. If Google has not yet indexed a URL, the analysis still runs on the fetched HTML but does not correlate against zero GSC data.
• Caps. The flow refuses with a localized message that names the metric (content_analyses) — buy a pack or wait for the reset.
• Provider degradation. Firecrawl or an AI provider outage surfaces as partial or refunded — never as a silent success.

Connecting the MCP surface

If you use Claude Code, Claude Desktop, Cursor, or VS Code, you can drive Content Intelligence from the RankMeFast MCP endpoint — see RankMeFast MCP.

Export and deletion

Every analysis is included in the account export. Deleting your account purges analyses inside the grace window documented in Account security. We never publish anything you have not explicitly accepted.