SEO and AI Search
Seven jobs: keyword research, technical audit, product page SEO, programmatic page drafts, a monthly AI visibility sweep across ChatGPT, Gemini, Perplexity and AI Overviews, citation ready content, and the monthly organic calendar that Content Publishing executes.
Who it is for
The owner who has never opened Search Console, and the marketer who wants to know whether AI answers mention the store at all, measured with a fixed prompt set, not a flattering one.
What it reads from the brain
Before every run. Sections are addressed as layer.section.
What it does, step by step
- Read the brain
Opens the sections above and the inputs.
- Gather evidence
Crawls the store, pulls keyword volumes and search results, runs the AI visibility sweep on the fixed prompt set for your category, and reads Search Console if connected. Every finding cites its source.
- Decide and draft
Ranks technical and on page findings, picks keyword targets with volume and intent, drafts schema, product page fields and up to ten programmatic pages a month, and writes the organic calendar. Drafts use your customers' vocabulary and only claims_allowed.
- Advise stops here
You get the audit, the keywords, the AI visibility report and the drafts.
- Act
Writes schema and metafields after approval, creates pages as drafts, and hands the calendar to the brain for Content Publishing to execute.
- Write back
What worked and the decisions log are appended.
Tools it uses
storefront.fetch, storefront.pageThe store as a crawler sees itseo.crawlTechnical auditseo.keywords.volumeVolume and intent per targetseo.serpWho ranks todayai_visibility.sweepMonthly sweep across ChatGPT, Gemini, Perplexity and AI Overviews on the fixed category prompt setgsc.performanceSearch Console, the only first party AI signal (the generative AI report)shopify.metafields.setApproval gated. Schema and SEO fieldsshopify.pages.create_draftAct: pages as drafts. Publishing is approval gated
The three modes for this agent
Read tools only. You get the ranked audit, keyword targets, the AI visibility report and the page and schema drafts.
Schema and metafields written after your approval. Pages created as drafts; publishing waits for a tap. The calendar is handed to the brain.
Monthly on the 1st at 04:00. Same gates. The sweep runs, the fixes land as drafts for one tap.
Guardrails and approvals
Enforced in code. Items with a filled circle wait for your tap in Telegram.
- Content changes go to draft only.
- Never publishes a claim outside claims_allowed.
- Programmatic pages capped at 10 per month.
- The Search Console generative AI report is the only first party AI signal.
- The AI visibility prompt set is fixed per category and never customer written, to avoid the branded prompt trap that fakes share of voice.
- Every recommendation cites its evidence. Uncited is removed.
What it writes back to the brain
Dated and sourced. Learnings and performance sections are append only.
Example output
Example output. Structure only. Counts are placeholders.
SEO and AI Search, your store, monthly
Ran at data level 1 (storefront only). Connecting Search Console would add first party impressions and the generative AI report.
- Technical findings ranked: [n], top three in audit.csv with the URL for each.
- Keyword targets: [n], with volume and intent, in keywords.csv.
- AI visibility: your store was cited in [n] of [N] fixed category prompts across ChatGPT, Gemini, Perplexity and AI Overviews. Planned: competitors cited more often listed by name in ai-visibility.md.
- Drafted: schema for [n] product pages, [n] programmatic pages (cap 10 a month), and the organic calendar for next month, handed to the brain.
Waiting for your approval to write schema and metafields on [n] products. Pages are drafts.
ai-visibility.md (structure)
| Prompt (fixed, category) | Engine | Your store cited | Who was cited |
|---|---|---|---|
| [prompt 01] | ChatGPT | no | [competitor names] |
| [prompt 01] | Perplexity | yes | [competitor names] |
Also delivered: audit.csv, keywords.csv, drafts.md.