Documentation

Workflow

From measurement to published improvement

An end-to-end workflow: prompt gap → page diagnosis → copywriting → verification.

8 min read

The loop GEO teams should run

Monitor prompts → Diagnose gap stage → Fix the right page → Re-evaluate → Re-monitor

Lumar GEO Studio supports each step in one product. This is the practitioner workflow — not a feature tour.

1. Build a representative prompt set

Goal: Sample the intents that matter commercially, not every imaginable query.

  • Group prompts into topics that match how your team plans content (pricing, integrations, comparisons, category education)
  • Include competitive and category-level prompts, not only branded queries
  • Set a schedule cadence that matches how fast your market moves — daily for volatile categories, less often for stable ones
  • Connect Google Search Console in project settings when available — it strengthens search-demand signals on page eval

2. Monitor at the right altitude

ViewUse it for
Topic overviewAllocate team time — which themes need work
Prompt detailDiagnose why a specific query is won or lost
Answer logsQualitative read — tone, named brands, cited URLs
Cited pagesWhich URLs compete for citations in a topic
Brand mentionsPositioning signal when citations are lagging

Do not flatten everything to one visibility number. Split citation vs mention diagnostics first.

3. Route issues to the right fix

Use the diagnostic playbooks:

  • Pipeline upstream (never cited, wrong page type) → page strategy, new URL, IA change
  • Page fitness (cited or candidate URL with weak eval) → copywriting workspace
  • Mention-only presence → quotable proof blocks, comparison tables, definition paragraphs

Enable auto-crawl for cited pages in project settings so newly discovered URLs enter evaluation without manual setup.

4. Run a copywriting session (page-level fix)

When page scores or cards identify fixable structure:

  1. Open the URL in Copywriting
  2. Generate reviewed plan and cards — deterministic eval → strategist review → pinned queue
  3. Read the strategy brief (north star, pillars, pinned tasks) before accepting drafts
  4. Work pinned cards in order — click a card to anchor chat to the right section
  5. Accept only copy you would publish; Reject and iterate
  6. Re-evaluate after a meaningful batch of accepts — scores lag until you do

The workspace is editorial, not autonomous. Cards flag missing proof — supply approved facts; do not treat draft text as verified.

Metric definitions: Metrics Glossary

5. Verify at both layers

After publishing live changes:

  • Page-level: Re-eval in copywriting confirms Precision, Recall, and Evergreen Health moved
  • Prompt-level: Wait for scheduled prompt runs — generative behaviour changes on its own timeline

An improved Evergreen Health score does not instantaneously lift visibility. Track prompt trends over the next cadence cycle.

6. Close the loop in reporting

Document for stakeholders:

  • Prompt/topic affected and business intent
  • Pipeline stage diagnosed (selection vs chunk vs mention)
  • URL changed and headline metric movement
  • Prompt visibility trend over subsequent runs

That narrative connects GEO methodology to outcomes — the difference between "we edited a page" and "we fixed why we lose comparison queries."

Quick start (new to Lumar GEO Studio)

  1. Create a project with brand domain and AI platforms
  2. Add topics and prompts (generate or manual)
  3. After first runs complete, open the weakest strategic topic
  4. Follow the loop above

For account setup steps, use Start from the homepage or your account team's onboarding guide.