Skip to content

Search evidence and choose the next action

This page is for the moment after a manager asks, “What do we know, and what should we do next?” Foodism Claw is most useful when it searches evidence, shows citations, and turns that into one practical move for the store.

Restaurant operator searching Foodism Claw evidence

The page is not a generic chat box. It is an operator workflow that combines live retrieval with a maintained merchant knowledge file.

LayerWhat it pulls fromWhy it matters to operators
Live evidence retrievalDiagnosis findings, Google profile data, reviews, uploaded menus, trade-area reports, and agent deliverablesHelps answer today’s question with current facts
Merchant wiki contextsoul.md and enriched merchant profile notesKeeps stable business context, signature dishes, positioning, and known facts in one place
Cited answer cardsSource references and preview links in the workbenchLets the operator check the original evidence before acting
Deliverable previewGenerated plans, drafts, or reports in the right panelMakes it easy to move from answer to execution

In practice this is a hybrid knowledge flow: fresh retrieval for changing signals, plus a persistent merchant wiki for accumulated context. Andrej Karpathy’s LLM Wiki note describes the same idea at a higher level: do not rediscover the business from scratch on every question, keep a maintained knowledge layer between raw sources and answers.

Ask like an operator, not like a slogan writer

Section titled “Ask like an operator, not like a slogan writer”

Good questions are short, evidence-seeking, and tied to a decision:

GoalAsk Foodism ClawWhy this works
Monday priority”If I have 15 minutes, what should I fix first today?”Forces one ranked action instead of a long brainstorm
Review operations”Which reviews are safe to reply to now, and which need owner review?”Separates routine replies from sensitive cases
Local SEO”Which missing profile field is most likely hurting search or calls?”Connects profile completeness to visibility and conversion
GEO content”What question are guests or AI answer engines likely to ask about this store?”Helps prepare content that is clear in Google and AI-generated answer surfaces
Menu clarity”Which dishes or menu facts are mentioned often enough to feature this week?”Grounds content in repeated evidence, not guesswork
Competitor watch”Which nearby search terms or competitors deserve attention this week?”Moves from generic SEO talk to local market action
Website check”What should we update on the website before we share it?”Catches mismatch between public facts and landing pages

Weak questions are too vague:

  • “Make us popular.”
  • “Do marketing.”
  • “Tell me everything.”

Better: “Which one issue is most likely to affect visits, trust, or orders this week?”

When QA/Search is working well, each answer combines changing signals with slower-moving store context.

Source typeExample use in QA/Search
Diagnosis findingsPrioritize missing hours, weak photos, broken links, or low-trust profile gaps
Google reviewsFind repeated praise, complaint themes, and reply candidates
Merchant wiki (soul.md)Reuse stable facts such as signature dishes, audience, tone, and brand framing
Trade-area or SEO reportsCheck search demand, nearby competition, and map visibility patterns
Uploaded filesRead menu PDFs, owner notes, event briefs, or promo documents
Agent deliverablesReuse plans, summaries, and drafts instead of asking the same question again

This matters for SEO and GEO because restaurant answers often need both:

  • current signals, such as today’s reviews or missing profile fields
  • compiled context, such as what the restaurant really sells and how it should describe itself consistently

How to judge whether an answer is safe to use

Section titled “How to judge whether an answer is safe to use”

Before acting, check four things:

  • Does it show which sources it used?
  • Does it separate fact from suggestion?
  • Does it lead to one practical next step?
  • Does it ask for human confirmation when price, address, hours, offers, or sensitive reviews are involved?

If the answer feels generic, narrow the scope:

  • “Use only the diagnosis result.”
  • “Use only recent Google reviews.”
  • “Compare reviews with soul.md before answering.”
  • “Separate today’s action from this month’s content ideas.”
  • “Write this as a task for the right assistant.”

For restaurant teams, SEO and GEO are not abstract channels. They are questions guests ask before they visit, order, or trust the place.

Operating needGood QA/Search prompt
Google ranking support”Which missing public facts should we fix first to improve search trust?”
AI answer readiness”If someone asks what this restaurant is known for, what can we answer from evidence?”
Review-driven content”Which review themes are strong enough to turn into a post or website headline?”
Competitor awareness”What nearby competitor pattern should we react to this week?”
Website consistency”Where do the profile, reviews, and website tell different stories?”

Use QA/Search to decide. Then hand work off when the answer is clear.

If the answer says…Next move
The profile has missing or wrong factsStart Google Store Decoration
Reviews need replies or triageStart Google Review Ops
There is a clear local content angleStart SEO Post Generator
The issue is neighborhood, competitor, or search demandStart Google Trade Area Analysis or Google SEO Audit
The website is incomplete or mismatchedStart Dineway Site Builder or a website update task

QA/Search is valuable because it reduces repeated guesswork. Ask once, inspect the evidence, then either act yourself or start the right assistant with a narrower brief.