aio

Prompt file imported from GAIA-Health/gaia-health.github.io (.claude/commands/aio.md). Copyright stays with the author.

AIO / GEO Context Loader for Go Go Gaia

You are starting (or continuing) AI-search optimization work (ChatGPT, Google AI Overviews, Perplexity, Gemini, Copilot). Load the standing strategy, context, and current status, then give the user a tight briefing. Do NOT ship content or strategy changes unless the user explicitly asks.

1. Read the reference docs (in this order)

1. docs/AIO.md — the primary AIO/GEO strategy: current state, ranked levers, goals, the core citation principle, the playbook, measurement approach, guardrails, backlog, and the dated Research Log. This is the source of truth.
2. docs/AIO-CLIFF-MAY2026.md — the May 21 2026 ChatGPT cliff incident: root cause (GPT-5.5 rollout, industry-wide, not us) + per-platform breakdown.
3. docs/AIO-SCENARIO-MODEL.md — floor/base/ceiling math + the TAM check (3k/mo is citation-share/CTR-bound, not TAM-bound).
4. docs/GROWTH-MODEL.md — how AIO ladders into the master 5k MAU model.
5. docs/DIFFERENTIATION-AND-CPP-COPY.md — source of truth for app claims; every liftable differentiator must match it.
6. Skim docs/SEO.md — the sibling channel. Google AI Overview wins surface partly through Google ranking, so the two strategies interlock.

If any are missing, say so and continue with what's available.

2. Know the tooling

• GA4 (property 443880004, credentials .ga-credentials.json in repo root, gitignored): pull AI traffic by source. Always filter platform=web — the property mixes iOS app and web streams. Pattern: npx --yes google-analytics-cli report --property 443880004 --credentials .ga-credentials.json --dimensions "sessionSource" --metrics "totalUsers,sessions" --date-ranges '[{"startDate":"30daysAgo","endDate":"today"}]' --dimension-filter '{"filter":{"fieldName":"platform","stringFilter":{"value":"web"}}}'
• Measurement trap: Google AI Overview citations mostly land in google / organic, NOT a separate AI source. Only Gemini chat referrals (gemini.google.com) break out cleanly. Never conclude AIO is failing from the AI-source number alone.
• GSC is under a different email — ask the user for CSV exports; don't try to pull it.
• Manual prompt audits (monthly): run "best pregnancy app 2026" / "best period tracker app" in ChatGPT, Google AI Mode, Perplexity, Gemini; record if we're named, how described, and who gets cited instead.
• For focused page-level analysis, the /seo-geo skill covers crawler access, llms.txt, citability scoring, and platform-specific checks.

3. Optionally pull live state (read-only, only if it helps)

If the user wants current numbers, run the GA4 AI-source pull above and compare against the per-platform baseline in the cliff doc (post-cliff corrected baseline: ~110 AI users/mo).

4. Report (keep it tight)

Give the user:

1. Strategy spine in one line: specific, verifiable facts win AI citations — self-praise does not; lean weight toward Google AIO/Gemini + Perplexity (controllable, insulated) while attempting ChatGPT recovery as upside.
2. Current state — post-cliff baseline (~110/mo), base case (~600–800/mo by Aug 2026), 3k = ceiling/north-star not the plan. Note anything newer from the Research Log.
3. Ranked levers — from AIO.md (Google AIO/Gemini first, ChatGPT re-citation, Perplexity, off-site authority, content surface area).
4. Open backlog — pull from AIO.md's "Open questions / backlog" checklist.
5. Guardrails — no superlatives (they also lose citations), no medical/efficacy claims, every feature claim verified against the differentiation doc, "free" is a fact not the lead, be generous to competitors.

5. Then ask what they want to do

Common next steps: run the monthly prompt audit across all 4 engines, audit verdict-block coverage on comparison pages, ship a vs-content gap page (doubles as citation fuel — see docs/SEO.md), work the off-site authority backlog (Wikipedia, Reddit), or re-pull the GA4 per-platform numbers.

After any session that produces findings, append a dated entry to the Research Log in docs/AIO.md; when a finding changes strategy, also update the section above it so the doc stays the source of truth.