hypd_search_help
ChatGPTSearches HYPD's help center for articles that answer a question about HYPD itself: connecting or reconnecting ad accounts, plans, billing and credits, privacy and data access, what HYPD can do, and setting HYPD up in Claude, ChatGPT or other AI clients. Returns the matching articles' text and links; answer from them and cite them. Not for advertising advice or the user's account data; use the platform tools for those.
hypd_show_getting_started
ChatGPTShows the user's getting started steps for HYPD SEO & GEO: connecting Search Console with a selected property, a first look at which queries and pages lost organic clicks, and checking whether ChatGPT names their brand. Use it when init or seo_gsc_list_sites says so, or when the user is new or asks how to get started; a concrete SEO question goes straight to the seo_* tools. Clients that render MCP Apps show the steps with a button each: a click sends that step's prompt as the user's next message, or opens the HYPD app. There, say only that the steps are shown above: do not repeat them and do not run a step before the user clicks it. Elsewhere, list the steps with their prompts and links.
init
ChatGPTReturns the SEO & GEO guide — call it first in a new conversation. It explains which seo_* tool answers which question, workflows for an organic traffic drop, a content brief, a competitor gap and an AI visibility check, which data is the user's own and which is a third-party estimate, how lookups use the plan allowance, and starter prompts.
seo_ai_visibility
ChatGPTAsks ChatGPT a buyer's question with web search on and returns its answer, the pages it cited and the searches it ran: the GEO check for ChatGPT. Pass brand to see whether the answer names it, and domain to see whether that site is among the cited sources; run it for the user and a competitor to compare. Write prompt the way a buyer would, in their language, and set country to their market. Answers vary: runs (1–3) repeats the question and the summary counts mentions and citations across runs. Citations aren't guaranteed even with web search. A slow, costly call: one question per call, few runs. Returns a row per run, a summary and a result_id.
seo_backlinks_anchors
ChatGPTThe anchor texts other sites use when they link to a domain or one page, most backlinks first: each with its backlinks, referring domains, followed share and kind (branded, url, empty for image links, or other words). The summary weighs the mix of kinds, a quick read on how natural the profile looks. Use it to compare a competitor's link wording with the user's, or for anchor themes in a link-building plan. 50 anchors by default, up to 500; more cost more. Returns rows and a result_id; paging with result_id is free.
seo_backlinks_referring_domains
ChatGPTThe domains linking to a domain or one page, strongest first (or by most backlinks with sort_by): each with its link authority rank, backlinks to the target, followed share, spam score and first link date; the summary gives the total. Use it for link-building targets that link to a competitor but not to the user, or with status "lost" to see who stopped linking. 50 domains by default, up to 500; more cost more. Returns rows and a result_id; paging with result_id is free.
seo_backlinks_summary
ChatGPTThe links pointing at a domain or one page, in numbers: link authority rank (0–100), backlinks, referring domains, the followed share, spam score, broken backlinks, when links were first seen, and the spread by top-level domain, link type, platform and country. Use it as the authority side of a competitive brief, or to judge whether a page can compete for a keyword; call it once per site to compare the user's with a competitor's. For who links use seo_backlinks_referring_domains; for the link wording seo_backlinks_anchors. Returns one row and a result_id.
seo_competitor_domain_overview
ChatGPTA site's organic and paid Google footprint in one market: keywords ranked, how many in the top 3 and top 10, estimated monthly traffic and its value, and keywords gained and lost. Compare the user's site with competitors in one call by passing domains (up to 50) instead of domain. Domains go without https:// or a path ("acme.com"), with location and language for the market; ask if unknown. Third-party estimates refreshed weekly, not the user's Search Console data. Returns rows and a result_id. For the keywords behind the numbers use seo_competitor_ranked_keywords.
seo_competitor_keyword_gap
ChatGPTCompares the Google organic keywords of the user's domain and one competitor in one market. mode "gap" (default) lists keywords the competitor ranks for and the user doesn't, highest volume first: the content to create. mode "shared" lists keywords both rank for, with both positions and URLs: where to win ground. Rows carry volume, difficulty, intent and CPC. Pass domain and competitor without https:// or a path, plus location and language; ask if unknown. Filter with min_volume. Third-party estimates. Returns rows and a result_id; paging with result_id is free. More rows cost more.
seo_competitor_keywords_for_site
ChatGPTKeywords relevant to a domain or page, ranking or not, highest volume first, with difficulty, intent, CPC and trend. On the user's own site it suggests topics to cover; on a competitor's, the topics their content targets. Filter with min_volume and max_difficulty. Pass target as a bare domain or a page URL with https://, plus location and language; ask if unknown. For what a site already ranks for use seo_competitor_ranked_keywords. Third-party estimates. Returns rows and a result_id; paging with result_id is free. More rows cost more.
seo_competitor_ranked_keywords
ChatGPTThe Google keywords a domain, or one page, ranks for in one market: position, ranking URL, estimated traffic, volume, difficulty and intent, most traffic first. Pass a page URL with https:// as target for that page alone, to find a competing page's content gaps against the user's. Filter with min_volume, max_position (10 = page one) and branded ("exclude" for non-branded demand). Pass location and language; ask if unknown. Third-party estimates; for the user's own measured rankings use seo_gsc_query_search_analytics. Returns rows, a summary with the total count and a result_id; paging is free. More rows cost more, so keep limit modest.
seo_competitors
ChatGPTThe sites that compete with a domain in Google's organic results, one row each: keywords shared and the average position on them, then the competitor's organic and paid keyword counts, estimated monthly visits and what that traffic would cost to buy in USD. Skip large platforms such as youtube.com or reddit.com, which share keywords with almost any site. Pass domain without https:// or a path ("acme.com"), and location and language for the market. Third-party estimates. Results keep 24 hours: result_id with page and page_size re-reads them for free. Then use seo_competitor_keyword_gap on the closest ones.
seo_ga4_landing_pages
ChatGPTWhat happens after the click, per landing page: sessions, users, engagement rate, average session duration, conversions and revenue by landing page (with its query string), most sessions first. To see organic search only, pass a dimension_filter on sessionDefaultChannelGroup equal to "Organic Search". Pairs with seo_gsc_query_search_analytics: Search Console says which pages win clicks, this says which of them engage and convert. Pass property_id from seo_ga4_list_properties and date_ranges; optionally limit. Dimensions and metrics are fixed.
seo_ga4_list_properties
ChatGPTThe user's GA4 properties, grouped by Google Analytics account, as HYPD last synced them: the property_id (properties/<digits>) seo_ga4_landing_pages and seo_ga4_run_report take. Call it before any organic traffic question that needs sessions, engagement or conversions rather than clicks. Takes no arguments. GA4 is optional on this surface: an empty list means it isn't connected, and Search Console and the research tools still work, so carry on with those and mention the user can connect GA4 in HYPD.
seo_ga4_run_report
ChatGPTAn open GA4 report on the user's property for questions seo_ga4_landing_pages doesn't cover: any dimensions and metrics over up to 4 date_ranges, with dimension_filter or metric_filter (one condition on a field, or an andGroup or orGroup of them), order_bys, limit and offset. For organic traffic, filter sessionDefaultChannelGroup to "Organic Search". GA4 allows at most 10 metrics and 9 dimensions per request. On a rate-limit error, run GA4 calls one at a time and retry after the wait it gives. Pass property_id from seo_ga4_list_properties.
seo_google_trends
ChatGPTHow search interest in 1–5 terms moves over time on Google web search, news, YouTube, images or shopping: the series, the regions where interest is highest and, for a single term, related topics and queries, top and rising. The summary gives each term's direction with its % change, its peak, and the peak and low months over two years or more. Use it for "is this topic growing?", content timing and seasonality. Values are relative interest (0–100 within one call), not volume: use seo_keyword_search_volume for volume. Pass location, or omit it for worldwide. Returns rows and a result_id; paging with result_id is free.
seo_gsc_compare_periods
ChatGPTCompares two periods of the user's organic search performance with correct date math: pass the current period, and the previous one defaults to the equal window just before it. One row per page (default) or other dimension, with current, previous and delta clicks, impressions, CTR and position: the tool for "organic traffic dropped: since when, and which pages and queries lost it?". Keep data_state "final" (the default), or unfinalized recent days read as a drop. current_truncated or previous_truncated means a side was cut: raise row_limit before calling a row new or lost. Position deltas compare impression-weighted averages, not ranks.
seo_gsc_inspect_url
ChatGPTGoogle's index verdict for one URL of the user's property: coverage state, Google-selected versus declared canonical, last crawl, the sitemaps that list it, and mobile and rich-result checks. The tool for "why doesn't this page rank or show up at all?". inspection_url must belong to site_url's property. One URL per call: for several, call it once per URL in parallel (the per-site quota is roomy).
seo_gsc_list_sitemaps
ChatGPTThe sitemaps submitted for one of the user's Search Console properties, with when each was last submitted and downloaded, whether it is pending, and its warning and error counts by content type. Use it for crawl and indexing hygiene: a sitemap with errors, or important pages missing from every sitemap, often explains pages Google hasn't indexed. Copy site_url from seo_gsc_list_sites.
seo_gsc_list_sites
ChatGPTThe Search Console properties the user has selected in HYPD, each with its permission level: the site_url every other seo_gsc_* tool takes. Call it first for any question about the user's own organic performance and copy site_url exactly as returned ('sc-domain:example.com' is a domain property, 'https://example.com/' a URL-prefix one; never turn one into the other). Takes no arguments. An empty list means Search Console isn't connected yet or no property is selected: the research tools still work without it, so offer those and hypd_show_getting_started. A property added in Google lately appears once the user refreshes Search Console in HYPD.
seo_gsc_page_query_breakdown
ChatGPTPass exactly one of page (a URL) for the queries it gets impressions and clicks for, or query for the pages Google shows for it. The tool for title and meta rewrites, content refreshes and cannibalisation checks (several pages competing for one query), done with a filter so it never trips Google's load quota. Rows carry clicks, impressions, CTR and impression-weighted position; truncated with row_count says more rows existed. A page's rare queries are anonymised, so its query rows never sum to its totals. Copy site_url from seo_gsc_list_sites.
seo_gsc_query_search_analytics
ChatGPTThe user's own organic clicks, impressions, CTR and average position from Search Console, grouped by up to two dimensions (query, page, country, device, date, search appearance) with filters. Position is impression-weighted, not a rank: filter position bands after fetching. To see a page's queries or a query's pages, use seo_gsc_page_query_breakdown instead of grouping by both (Google's load quota). Dates default to the 28 days ending 3 days ago; data_state 'final' is the default, 'all' adds fresh but provisional days. row_limit defaults to 250 (max 25,000); page with start_row. truncated and row_count flag an incomplete result. Rare queries are anonymised, so query rows never sum to the site's totals. Copy site_url from seo_gsc_list_sites.
seo_keyword_ideas
ChatGPTBroad keyword ideas from 1–20 seed keywords, with monthly search volume, competition and bid ranges, from Google Keyword Planner when the user has Google Ads connected and from a third party otherwise. Use it to brainstorm a topic's keyword space; for SEO difficulty and intent use seo_keyword_overview, for adjacent terms seo_related_keywords, for long-tail variants seo_keyword_suggestions. Pass location and language for the market (United States and English when omitted). Results keep 24 hours: result_id with page, page_size and sort_by re-reads them for free, without keywords, location or language.
seo_keyword_overview
ChatGPTScores up to 100 keywords for organic search in one market: keyword difficulty (0–100, how hard page one is to reach), main and secondary intent (informational, navigational, commercial, transactional), monthly volume, CPC, yearly trend and the SERP features Google shows. Use it to pick a page's target keyword or decide whether a term is worth the effort. Pass location and language from the market the user named; ask if unknown. Keywords without data are left out. Returns rows and a result_id; paging with result_id is free.
seo_keyword_search_volume
ChatGPTMonthly search volume, the 12-month trend, CPC and competition for 1–1,000 keywords you already have, from Google Keyword Planner when the user has Google Ads connected and from a third party otherwise. Use it to size a keyword list; seo_keyword_overview adds difficulty and intent for up to 100. Pass location and language for the market (United States and English when omitted). Results keep 24 hours: result_id with page, page_size and sort_by re-reads them for free.
seo_keyword_suggestions
ChatGPTLonger searches that contain a seed keyword ("running shoes" → "best running shoes for flat feet"), highest volume first, each with difficulty, intent, CPC and trend: the place to find low-difficulty long-tail targets and FAQ-style questions around a topic. Filter with min_volume and max_difficulty. For adjacent terms without the seed use seo_related_keywords. Pass location and language for the market; ask if unknown. Returns rows and a result_id; paging with result_id is free. More rows cost more.
seo_page_audit
ChatGPTOpens one URL in a real browser and returns what Google's renderer would see: the JavaScript-rendered DOM, HTTP status, redirect chain and server timings. Use it for on-page SEO checks a plain fetch misses: the title tag, meta description, canonical, robots meta, hreflang, structured data and headings as rendered. Pass the full url with https://; browser_preset picks desktop (default), mobile or tablet. For speed and Core Web Vitals use seo_page_lighthouse; for the content outline seo_page_outline.
seo_page_lighthouse
ChatGPTRuns Google Lighthouse on one URL: Performance, Accessibility and Best Practices scores with the metrics behind them, Core Web Vitals included, the page-experience side of ranking. Pass the full url with https://; for_mobile true runs the mobile audit, which is what Google indexes with (desktop by default). categories narrows it to any of performance, accessibility and best_practices. For how the page looks use seo_page_screenshot.
seo_page_outline
ChatGPTA page's main content as an outline, without header, footer or sidebars: each heading with its level and the paragraphs, lists and tables under it. Sections have stable ids ("h2-3" is the third H2) and their heading path, so a content brief can say exactly where an edit goes. The summary has the title, word count, heading counts and heading problems (no H1, several H1s, skipped levels). Use it for content briefs and to compare a competitor's page structure with the user's. Pass the full url; javascript true renders scripts first and costs more, for pages whose outline comes back empty. Returns rows and a result_id; paging with result_id is free.
seo_page_resources
ChatGPTLists the scripts, stylesheets and images one URL loads, with their load timing: the way to find what slows a page down, broken resources (resource_type "broken"), or whether analytics and tag manager scripts load at all. Pass the full url with https://; resource_type narrows the list to script, stylesheet, image or broken.
seo_page_screenshot
ChatGPTCaptures one URL in a real browser and returns the image inline: what a searcher sees on arrival, above the fold and on mobile. Use it to judge whether a page answers the query it ranks for, or to compare a competitor's page with the user's. Pass the full url with https://; full_page captures the whole scrolling page; browser_preset picks desktop (default), mobile or tablet. Judge the page from the image itself.
seo_related_keywords
ChatGPTExpands one seed keyword through Google's related searches into adjacent and secondary terms for a content brief, highest volume first, each with difficulty, intent, CPC and trend. depth 1 (default) gives the close neighbours (about 8), depth 2 one step further (about 70). Filter with min_volume and max_difficulty. For longer searches that contain the seed use seo_keyword_suggestions. Pass location and language for the market; ask if unknown. Returns rows and a result_id; paging with result_id is free. More rows cost more.
seo_serp_results
ChatGPTGoogle's results page for one keyword in one market, by block: organic, featured_snippet, people_also_ask, ai_overview, local_pack, related_searches and more. Pass item_types for the blocks you need. Add "ai_overview" for Google's AI Overview text and the pages it cites, with cited_domain to check whether the user's site is among them: the GEO check for Google. With "people_also_ask", paa_depth (1–4) opens the questions' answers and sources for an FAQ brief. Pass location and language from the market the user named; never assume worldwide. Returns paged items and a result_id; paging with result_id is free. Google data combined with third-party data.