ローカル検索の機会調査

# Local Search Opportunity Research

## Loadout capabilities

Use <action-tag>api:tinyfish_api:fetch_urls</action-tag> to read the business's public home, location, menu, and service pages. Use <action-tag>api:tinyfish_api:search</action-tag> for a bounded sample of public search results and customer-language variations in the stated location. Inspect current schemas, authentication, and cost before execution. Search snippets are evidence of language and competing pages at a specific time, not a measured ranking baseline.

For a WordPress site, use <integration-tag>api:wordpress_api</integration-tag> to identify the connected post-inventory route and setup requirements for a later follow-up. Its current Action reads published posts only after the merchant configures that site's REST base URL and API credential. This Skill does not execute that connected Action; it uses public page fetches and search to identify content gaps. The WordPress Integration does not list pages, edit content, or expose SEO-plugin metadata; keep the no-sign-in public workflow available when no connection exists.

Optional paid expansion: use <action-tag>mcp:ahrefs_mcp:site_explorer_organic_keywords</action-tag> for a small, country-filtered domain ranking sample. Ahrefs is Aident-managed where the live Action and Vault permit it; this does not require the merchant's Ahrefs sign-in. Its variable API-unit pricing requires a cost preflight and a user-approved spend cap before any paid query. The default workflow ends after public-site and public-search research. If Ahrefs is unavailable or unaffordable, omit volume and difficulty rather than inventing them. Follow the live Ahrefs doc Action for its current input schema and use `mode=subdomains` for domain analysis.

## Inputs and sampling

Require the public domain, physical locations or service areas, primary offerings, and target language. Ask for a maximum research spend only if the user wants Ahrefs expansion. Use one country and the specific city or neighborhood for each query; never treat country-wide demand as hyperlocal demand. Start with at most three location pages, five offerings, and ten public search queries across brand, category, offering, and customer-question intent. Exclude unrelated cities, businesses with the same name, and terms for unverified offerings. State query, locale, date, and sample size for each observation.

## Workflow

1. Extract verified entities from the site: location names, offered products or services, ordering or booking methods, and local differentiators. Mark an offering uncertain if it appears only in a stale snippet or third-party directory.
2. Generate candidate phrases from customer needs: `[category] [city]`, `[offering] near [neighborhood]`, `where/how/can I [need]`, and brand or branch queries. Keep natural phrasing and language variants; do not fabricate search volume.
3. Search a bounded sample of these phrases. Record the result URLs, visible titles and snippets, and whether the business has a directly relevant page. A search result is an observation at one time and location, not a stable rank claim. Fetch a few directly relevant competitor pages only when needed to understand search intent; avoid copying their text.
4. If a spend cap was authorized, preflight the Ahrefs Action with the exact live schema and a small result limit. Stop when the cap would be exceeded, a quote is uncertain, or the Action requires a personal account. Label Ahrefs volume and traffic as third-party estimates. This Action does not necessarily return keyword difficulty. Do not equate estimated traffic with site analytics, nor interpret missing Ahrefs rows as zero demand.
5. Group candidates by location and intent: discover a place, choose a product or service, resolve a specific question, or navigate to the brand. Map each group to an existing page, a concrete improvement, or a justified new page. Avoid thin city-variant pages, unsupported superlatives, and unrelated keywords.
6. Prioritize by merchant relevance, evidence of demand, current page gap, conversion value, and execution effort. State confidence explicitly. If metrics are unavailable, rank qualitatively using the observed public evidence and explain uncertainty.

## Required output

Deliver three sections in this order:

1. **Evidence report:** public site and search URLs, query/location/date, what each source shows, and measured versus third-party estimated versus inferred fields. Include coverage gaps and conflicts.
2. **Prioritized strategy:** ranked keyword and question clusters with location, intent, evidence, target page or gap, confidence, and why that cluster outranks the next. Include brand terms separately so they do not crowd out non-brand opportunities.
3. **Operable actions:** for each top cluster, specify the exact existing or proposed URL, page title or section brief, merchant facts needed, owner, approval gate for any write, verification step, and expected observable result. A draft can be created without sign-in; publishing needs the site's authorized editor. Include a small monitoring plan using Search Console only after a merchant connects it, or public reruns with the same locale and queries.

For WordPress, map each action to the page or post URL and editor field: title, slug, H1, body section, internal links, and the SEO plugin's title, description, canonical, and index settings where installed. Include preview and rendered-page checks. Do not assume a plugin or use the read-only posts Action to write.

Never claim guaranteed ranking, featured snippets, AI citation, or a numerical traffic lift. If Ahrefs was used, report exact Action name, inputs, returned sample size, quoted and actual cost when available, and any omitted paths because of budget.