Spinlap Radar User Manual

Spinlap Radar is a comprehensive AI SEO Intelligence System designed to grow your brand's presence across the modern search landscape—including AI Answer Engines and Google Search.

Like platforms such as peec.ai, we focus on Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO): helping you measure how often engines like ChatGPT, Perplexity, Gemini, and Google AI Overviews cite your website, identifying visibility gaps, and scoring your content drafts to make them highly quote-worthy by AI search models.

AI Search & GEO (Generative Engine Optimization)

With zero-click search eating organic clicks, optimizing for traditional blue links is no longer enough. Your target audience is asking engines like ChatGPT and Perplexity complex questions, and these engines synthesize answers by quoting trusted, structured web references.

Spinlap Radar provides the tools to measure this visibility, score your drafts against AI citation signals, and maintain presence:

• AI Visibility / Share of Voice: We track how often your target questions mention or cite your brand domain versus your primary competitors.
• Citability Signals: The platform grades drafts on specific parameters that LLMs look for (such as definitional clarity, FAQ structures, fact schemas, and entities).
• AI Feedback Loop: As your published pages rank and get cited, the system correlates success back to our scoring parameters and automatically adjusts its content quality-gating weights.

Dashboard Guide (5-Step Pipeline)

Run the platform through the visual workspace dashboard in five steps:

1. Step 1: Set up client workspacesCreate isolated workspaces for each client or brand. Configure their target domain name, brand keywords, and API keys (such as Perplexity or DataForSEO credentials).
2. Step 2: Input AI Prompts & SEO KeywordsAdd the natural language queries that searchers ask AI engines regarding your category (e.g. "best automated invoice workflow software for startups"). Seed them automatically from top Google search keywords.
3. Step 3: Generate Citable Outlines & DraftsGenerate a page for a target topic. The engine analyzes live SERPs to build a structured content brief (outline, headings, word count target) and drafts an optimized copy designed to satisfy the query's search intent.
4. Step 4: Editorial & GEO Scoring FunnelBefore anything goes live, drafts pass through our 3-stage scoring funnel:
   • 🟢 Post (70+): High relevance and citability score. Ready to ship.
   • 🟡 Review (55-69): Suggests specific improvements (e.g. "Add a definitional list" or "Crisp up entity clarity").
   • 🔴 Regenerate (<55): High risk of generic AI phrasing. Rewrite before launching.
5. Step 5: Direct Publishing & Citation TrackingConnect your WordPress connections to push approved drafts directly. Spinlap Radar runs daily EventBridge crons to monitor new citations, updates your share-of-voice charts, and tracks Google positions.

Managing Workspaces

Spinlap Radar is designed multi-tenant from the data layer up:

• Complete Data Isolation: Every workspace isolates its keywords, AI prompts, page drafts, CMS connections, and performance logs. Workspace boundaries are enforced contextually at the server layer.
• Role-Based Access: Invite team members and clients with Owner, Manager, or Read-Only permissions. Read-Only clients can browse dashboards and visibility trends without access to generation or publish tools.

The scoring funnel in detail

Every draft passes three stages before it can be published, ordered cheapest-first so that an obviously weak draft is rejected without spending money on it.

1. Stage 1 — Heuristics (no API cost)Deterministic checks on the text itself: length against the target, readability, whether a call to action is present, and hashtag or formatting hygiene. Drafts scoring below the floor are pruned here and never reach a model.
2. Stage 2 — Embeddings (relevance and novelty)The draft is embedded and compared against your topic and against your recent published work. Topic distance catches a draft that has wandered off-brief. Novelty catches the more insidious failure: a draft that closely repeats something you already published. A near-duplicate is forced into review rather than published, because two pages saying the same thing compete with each other instead of adding up.
3. Stage 3 — Model judgeSurviving candidates are judged together rather than one at a time. Comparative judging is markedly more stable than asking for an absolute score per draft, which drifts between runs.

The weighted result maps to one of three recommendations, and the thresholds are configuration rather than code:

• Post — high confidence. Eligible for automatic publishing if you have enabled it.
• Review — plausible but flagged. The dashboard shows which signal held it back, so the fix is specific rather than “make it better”.
• Regenerate — below the floor. Cheaper to draft again than to edit.

Publishing & CMS connections

Approved drafts can be pushed straight to your site. Three platforms are supported today:

• WordPress — application-password auth against the REST API. The most common setup and the least configuration.
• Ghost — Admin API key. Publishing and updating are both supported; an update has to reconcile against the post’s current revision, which the connector handles for you.
• Webflow — needs slightly more setup than the others, because a Webflow collection has no fixed schema. You supply the collection and a map from our fields to your field slugs. Anything you leave unmapped is omitted rather than guessed, since an unrecognised slug causes Webflow to reject the whole item.

Every connection can run in a dry-run mode first, which executes the full path — auth, payload construction, validation — and stops before the write. Worth using once per new connection.

Client reports

Each workspace can produce a branded, self-contained report page for a reporting period, shareable by link without the recipient needing an account.

One design decision is worth knowing about, because it surprises people: a report is a frozen snapshot. Generating a report for the same period twice creates a second report rather than updating the first. This is deliberate — a link you have already sent to a client must keep showing the numbers they discussed with you, not silently change the week after.

Automation & Schedules

Schedules run automatically in the background across all workspaces on EventBridge (Lambda) or scheduler threads:

• Daily Generation: Generates the highest priority topic outline + page draft for each client workspace.
• Daily Visibility Scan: Scrapes Perplexity, Google AI Overviews, and Google SERPs to record citations, mentions, and blue-link positions.
• Weekly Site Audits: Crawls your registered sites to flag SEO warning/critical regressions.
• Weekly Recalibration: Correlates your scoring parameters to citation and rank outcomes, retuning the system weights automatically.

Limits & Configuration

• Weekly Spend Cap: SEO_GENERATION_WEEKLY_LIMIT caps page generation requests on a rolling-7-day basis (default 25) to avoid model cost wildruns.
• Credentials: To enable AI Search Visibility scans, verify your Perplexity API keys and/or DataForSEO credentials are saved in your settings. Unconfigured engines will gracefully no-op.
• Strict Tenancy: Data is strictly filtered at the database level. API keys and actions are only valid for the workspace token present in the request headers.

Troubleshooting

A visibility scan returned nothing

Almost always missing credentials. An engine with no API key configured is skipped rather than failed, so the scan completes successfully having checked nothing. Confirm your keys are saved in workspace settings, then re-run the scan by hand rather than waiting for the next scheduled one.

Generation stopped partway through the week

Check the weekly generation cap. It is a rolling seven-day limit that exists to stop a misconfigured schedule turning into an unbounded model bill, and it halts before making the API call rather than after. Raising it takes effect immediately.

Everything is stuck in “review”

Usually the novelty check catching genuine repetition. If your topic list is narrow, successive drafts converge on the same points and get flagged as near-duplicates of your own recent work. Broaden the topics rather than lowering the threshold — the check is generally right.

Publishing succeeds but nothing appears on the site

Check whether the connection is still in dry-run mode. Dry runs report success because the whole path did succeed; it simply stopped before the write.

Scheduled jobs ran twice

Only one instance anywhere should own the schedules. If you run both a long-lived server and the scheduled deployment, both will fire. Disable the in-process scheduler on the deployment that is not the designated owner.

Related reading

• Free AI visibility audit — see which engines cite you today, and which competitors are cited instead.
• How to get cited by ChatGPT, Perplexity and Google AI Overviews — Ranking earns a blue link. Getting cited puts you inside the answer — and the two are decided by different things.
• “Crawled – currently not indexed”: what Google is actually telling you — It reads like an error. It is actually a verdict — and the fix is almost never in your code.
• Measuring AI visibility without fooling yourself — Ask an engine the same question twice and you get two answers. Any metric that ignores that is measuring noise.
• All field notes — everything we publish on AI search optimization.