Turn workspace evidence into the next defensible SEO action
Before committing to a recommendation, read how priorities and evidence are assessed.
One bounded guide
A real goal lens
Decision evidence retained
Choose the question the queue should answer
The goal selector changes the decision request, not just the label above an unchanged list. Crawl Foundry rebuilds the guide for Balanced, Growth, Quick wins, Protect, or Technical. The selected lens can admit different action families, move prerequisites ahead of dependent work, and change the final order. The underlying source evidence remains in its product, even when an action does not enter the current guide.
Balanced keeps the broadest quarterly view. Growth emphasizes qualified-lead opportunities such as new or improved content. Quick wins uses a low-effort, immediate lens. Protect favors ranking loss and monitoring. Technical concentrates on crawl findings, internal links, and consolidation work. Severe blockers can remain visible across lenses because they can invalidate the rest of the plan.
Treat the guide as a dated decision snapshot
The header shows when the report was evaluated and which intelligence mode the current coverage supports. An active report also has an evidence fingerprint. Saving a measure or requesting paid action intelligence is tied to that exact identity so a request cannot silently use a different evidence set.
When the active identity no longer matches, Crawl Foundry requests a refresh. A last known good guide may remain readable during that transition, but its fingerprint is deliberately withheld from mutations and paid AI. Wait for the refresh to finish before saving, starting, or purchasing an interpretation of that recommendation.
Read the source ledger before the recommendation
| Source | What Action Center can currently use |
|---|---|
| Keyword Database and Keyword Research | Stored keyword identity, demand, intent, difficulty, workflow state, provenance, research candidacy, and selection evidence within projection limits. |
| Domain and Competitive Research | Stored domain strength, organic visibility, referring-domain summaries, competitor gaps, domain rank, and bounded SERP comparison evidence. |
| Position Tracking | Current and previous measured position, ranked URL, market and device context, and a bounded portion of history. |
| Search Console | Authorized clicks, impressions, CTR, and average-position facts when a stored Search Performance snapshot exists. A connected property without that snapshot contributes status, not invented performance rows. |
| Site Audit and internal links | Run completeness, issue severity, affected scope, and stored link opportunities. The projection is intentionally smaller than the full explorers. |
| Content execution and workspace profile | Stored plan and page state plus confirmed or suggested business context. Content destinations fall back to the Keyword Database while Content Plans is unavailable. |
Ready, partial, stale, and truncated describe different limits
A source row carries status, item count, capture time, truncation, and limitation codes. A ready source can still be a bounded projection rather than the complete source product. Partial means some usable evidence survived, not that every relevant field is present. Missing means the guide cannot rely on that source for the current decision.
Facts also carry their own freshness state. The default source windows differ because a domain snapshot, a tracking measurement, a crawl, and a research run age at different rates. Use the observation date on the fact, not the report timestamp alone, when recency affects the decision.
The queue is a bounded guide, not a dump of every candidate
Crawl Foundry first checks hard blockers, business fit, minimum evidence, rankability requirements, dependencies, conflicts, and duplicate practical actions. It then applies the selected goal, prioritizes urgent loss or blockers, and diversifies the remaining work across action families, lanes, and sources.
The current decision guide can contain up to 30 actions. At most two retain the Do now band, and up to five form the focus portfolio. Those caps stop one noisy issue family or one source from occupying the whole workspace. Suppressed candidates are diagnostic evidence, not deleted source records.
Use the four-column board as a situation filter
The Overview groups actionable work into Technical, Improve pages, New pages, and Rankings. Each column shows its count, the number of critical actions, and the first ranked action in that field. That last item is the current lead recommendation, not a claim that the page or issue is objectively the worst on the site.
Selecting a column filters the visible recommendation list. It does not recalculate the evidence, change status, or remove work from the other fields. Context questions and pure data-readiness prompts stay outside this actionable snapshot.
Separate engine order from personal work priority
The decision guide arrives in deterministic order. In the current interface, a saved work item marked High is brought forward first. The selected goal's local affinity is applied next, followed by the guide order. There is no user-facing switch that replaces this with an arbitrary impact, effort, or evidence sort.
Starting a walkthrough freezes that round. A background refresh cannot move the next card while someone is deciding. Leave the flow and return to Overview to work from a newly evaluated queue.
Resolve business fit instead of letting the model guess
The workspace profile can hold offerings, audiences, markets, outcomes, editorial topics, brands, exclusions, and explicit target overrides. The compact editor currently lets a user confirm offerings, markets, and exclusions directly. Optional paid synthesis can propose a broader profile, but its suggestions become current only after a person accepts them.
When a saved or tracked keyword has unresolved fit, the guide can admit one calibration question. Marking a target relevant or not relevant creates a new confirmed profile revision. Choosing Not my topic from an eligible focus card records the same kind of target feedback; Later and a plain skip only advance the current session.
Open one action and inspect the claim behind it
A focus card shows category, impact, known effort, a plain-language rationale, and up to three lead facts by default. The first disclosure adds a three-step family-specific implementation plan and a verification statement. More context can reveal up to six evidence rows by default, business-fit and rankability signals, limitations, data needs, and source links.
These display limits keep the decision readable. They are not the size of the underlying source. Open the destination in Keyword Database, Keyword Research, Domain Overview, Competitive Research, Position Tracking, Site Audit, or Settings before changing production work.
For a technical action, follow the issue and affected-URL evidence before choosing the fix.
Missing evidence becomes a named data need
A recommendation can ask for business confirmation, an owned-source refresh, time for another measurement, or a billable keyword or SERP bundle. The need states which fact is missing and the largest decision change the additional evidence is expected to make. That upper bound is a policy estimate, not a probability that the action will succeed.
Stored reads and user confirmation do not create a provider charge. A billable bundle opens the existing Keyword Database enrichment dialog with the affected keywords, up to the current workspace limit. Keyword data and SERP each keep their own quote and confirmation. Nothing is purchased merely because the Action Center identified the gap.
Request a deeper action plan only when it is worth the quote
Eligible recommendations can request optional AI action intelligence. The confirmation shows the current price and named evidence sources before the run starts. A matching completed result is reused unless a person explicitly refreshes it. A refresh is recommended when the evidence fingerprint, context contract, prompt version, or age threshold has changed.
The result contains an assessment, confidence label, ordered implementation steps, a verification method, and missing-data notes. It does not reprioritize the guide. The packet contains bounded candidate facts, related signals, goal and confirmed business labels. Raw Brain documents, page bodies, uploaded files, account data, cross-workspace data, personal data, and credentials are excluded. Secrets detected during sanitization block the request.
Action intelligence currently returns German for the German app locale and English for the other app locales. A failed run reports whether the reservation was charged, released, or still requires billing reconciliation.
Start, save, and skip have different consequences
| Action | What it currently does |
|---|---|
| Start | Saves the recommendation as In progress when possible, advances the frozen walkthrough, and opens the exact destination for implementation. |
| Save | Captures the candidate, evidence fingerprint, destination, source, engine priority, guide band, verification code, and available baseline evidence as a personal workspace item. |
| Not my topic | For an eligible keyword question, records a not-relevant target override and advances the session. The profile revision can change a later guide. |
| Later or Skip | Advances only the current walkthrough. It does not archive source evidence or create a durable snooze date. |
The Board preserves commitments separately from the live guide
Saved, In progress, and Completed are user-owned work states. High, Normal, and Low are personal priorities, initially derived from the candidate priority but editable later. Within a column, personal priority comes before stored order and update time.
A current badge means the candidate still exists in the active guide. Historical means the saved snapshot remains available even though the current guide no longer contains that candidate. The board is bounded per user and workspace: the current default admits up to 200 open items and shows up to 50 completed rows. Archived items leave the board but remain available through paginated history.
Completion records work done, not SEO impact proved
Marking a card Completed records the completion time and schedules an outcome review window based on the category. It does not assert that rankings improved, a crawl finding disappeared, or a page was indexed. The current Board also does not expose the later comparison as a success badge.
Verify in the owning source: run a comparable crawl for a technical issue, wait for another tracking or authorized search-performance observation for a ranking action, or confirm publication and indexability for content. Keep scope, market, device, URL, and observation window comparable before attributing the change to the completed work.
Read empty and refresh states literally
With no usable keyword or technical evidence, Action Center shows onboarding paths to Keyword Database and Site Audit. If sources are usable but the selected goal has no admitted action, switch the goal or review business context. An empty goal view does not mean the workspace has no data.
A retained guide during refresh is useful for reading, but it is not current authority for a new save or paid interpretation. A refresh failure should send the reviewer back to source coverage and the last evaluation time rather than encourage a decision from an identity the system itself no longer accepts.
Example: Protecting rankings after a drop
After a traffic drop, a team switches the goal to Protect. The guide moves urgent ranking losses and technical blockers to the top and hides growth ideas that can wait.
The strategist opens the lead action, reads its evidence and verification statement, and presses Start, which saves it as In progress and opens the exact page in Site Audit where the fix belongs.
Frequently asked questions
Which goals can the Action Center optimize for?
Balanced, Growth, Quick wins, Protect and Technical. The goal changes which action families are admitted and how they are ordered, not just the label.
Does completing an action prove SEO impact?
No. Completed records that work was done and schedules an outcome review. Whether rankings or crawl findings changed is checked later in the owning tool.
What does the optional AI action intelligence cost?
The confirmation shows the current price and the evidence sources before the run starts. A completed result for the same evidence is reused unless someone explicitly refreshes it.
How does the Action Center know my business?
From the workspace profile: offerings, markets and exclusions can be confirmed directly. Confirmed business fit keeps irrelevant actions out of the guide.