Audit Technical Seo

---
name: audit-technical-seo
description: When the user wants a technical SEO audit of a site. Crawls the site, reads status codes, indexing, broken links, redirects, canonicals, duplicate titles and on-page gaps, confirms them on the pages that earn traffic (with the user's Google Search Console when connected), and ranks the fixes by impact. Also use when the user mentions technical SEO, an SEO audit, an SEO health check, crawl our site, broken links and 404s, duplicate titles, noindex or canonical problems, redirect chains, indexing problems, or is anything stopping Google from indexing our pages. A fall in traffic goes to diagnose-traffic-drop, a domain change or redesign to create-migration-plan, AI crawlers and llms.txt to check-ai-crawler-access, an audit for a prospect to prepare-client-pitch. Editing the site is out of scope.
compatibility: Requires the Manifold MCP connector, signed in with a Manifold account that has credits. Without it the skill stops and tells the user how to connect.
---

# SEO audit

A technical audit answers one question: can Google crawl, index and understand the pages that matter? This skill crawls the site, weighs every issue by how many pages it hits and whether those pages earn traffic, and ends in a fix list ordered by impact, not by the crawler's count. The fixes go to the user's developer; new content goes to the SEO content skills.

## Connector check

This skill runs on the manifold MCP server. Before the first step, confirm that its tools are available: look for `seo_run_technical_crawl` and `seo_get_domain_overview` (hosts often add a prefix, for example `mcp__manifold__seo_run_technical_crawl`).

- If no manifold tool is there, stop and tell the user: "This needs the Manifold connector. If you installed the Manifold plugin, connect Manifold on the plugin's Connectors tab (Customize > Plugins in Claude), run `/mcp` in Claude Code, or run `codex mcp login manifold` in Codex. Otherwise add the remote MCP server `https://mcp.manifoldmcp.com/mcp` and sign in: in Claude or ChatGPT as a custom connector, in Claude Code with `claude mcp add --transport http manifold https://mcp.manifoldmcp.com/mcp` and then `/mcp`. Other clients: https://www.manifoldmcp.com/docs/clients." Do not answer from memory instead.
- Then follow the tool checks in the [Google search notes](../create-seo-plan/references/platforms/google.md#tools): the `seo_*` tools are required, the `console_*` tools optional.

## Product context

Before the first question, look for the product marketing context file: `.agents/product-marketing.md` (in older setups `.claude/product-marketing.md` or `product-marketing-context.md`). If it exists, read it and take every input it answers (the site, the Search Console property, the market, the keywords that matter) from it; ask only for what it lacks. If it does not exist and the job needs more than two answers about the business, offer [manifold-get-started](../manifold-get-started/SKILL.md) first.

## Inputs to settle first

- **Site**: the domain. If a subdomain matters (a blog or docs host), check afterwards that the crawl reached it in `sample_urls`; if it did not, crawl that host as its own `target`.
- **Size**: roughly how many pages the site has. `max_pages` should cover it, up to 1,000 per crawl today. Default: 500 for a small business site, 1,000 otherwise.
- **JavaScript**: whether the pages build their content in the browser (many React and Vue single-page apps do). Step 1 tests it.
- **Search Console**: optional. With the `console_*` tools, index status comes from Google itself; the [Google search notes](../create-seo-plan/references/platforms/google.md#search-console) say how to check. The audit runs in full without it.
- **Budget**: a default run costs about 5 + 30 = 35 credits for 1,000 pages, or 5 + 300 = 305 rendered; `seo_get_page` and `get_task` are free. Say so before starting; pass `max_credits` if the user gave a budget.

## Steps

1. **Size it and test rendering.** `seo_get_domain_overview` on the site (5 credits): `organic_keywords` and `top_pages` say which pages earn traffic, so the audit weighs issues by them. With Search Console, `console_get_search_analytics` with `dimensions: ["page"]` for the last 90 days (free) gives real clicks per page: weigh by those instead. Then `seo_get_page` on the homepage and one key page (free). If `word_count` is tiny (under about 100) or `h1` is empty on a page that shows full content in a browser, the content needs JavaScript: crawl with `render: true`, and flag that Google must render it too.
2. **Crawl.** `seo_run_technical_crawl` with the domain as `target` and `max_pages` from the inputs (3 credits per 100 pages, 30 per 100 with `render: true`; charged on `max_pages` requested, not crawled). It returns a `task_id`. Call `get_task` (free) after `poll_after_s`; while it answers `TaskPending`, wait that long again. If `pages_crawled` is far below the site's size, the crawler was blocked or the internal links do not reach the pages; if it equals `max_pages`, the audit is a sample.
3. **Read the result.** `status_codes` counts the crawled pages per HTTP status; `broken_links`, `non_indexable`, `duplicate_titles` and `duplicate_descriptions` are counts. `issues[]` lists each failed `check` with `pages` and up to 10 `sample_urls`. Sort the checks into five buckets:
   - Indexing: `is_5xx_code`, `is_4xx_code`, `is_broken`, `is_redirect`, `canonical_chain`, `recursive_canonical`, and the `non_indexable` count.
   - Duplication: `duplicate_title`, `duplicate_description`, `duplicate_content`.
   - On-page: `no_title`, `no_description`, `no_h1_tag`, `title_too_long`, `low_content_rate`.
   - Site structure: `is_orphan_page`, `is_link_relation_conflict`, `is_http`.
   - Speed and hygiene: `high_loading_time`, `large_page_size`, `no_image_alt`, `no_favicon`.
4. **Confirm on the pages that matter.** For every sample URL that is also a top page, `seo_get_page` (free): `status`, `canonical`, `robots_meta` and `final_url` confirm the issue on the live page. With Search Console: `console_inspect_url` on up to 10 key URLs for `coverage_state`, `google_canonical` against `user_canonical` and `last_crawl_at`, and `console_get_sitemaps` for sitemap errors (free). With Bing connected, `console_get_crawl_issues` lists the URLs Bing failed to crawl with their inbound links (free).
5. **Rank the fixes.** Order by what blocks indexing, then by pages affected, then by the traffic of those pages: 5xx and noindex or wrong canonicals on pages that earn traffic first; then broken internal links and redirect chains; then duplicates; then on-page gaps; hygiene last. One template fix often clears hundreds of pages: say so when the sample URLs share a pattern.
6. **Deliver** a table: issue, check name, pages affected, sample URLs, why it matters, the fix, owner (developer or writer), priority. Put the `onpage_score` and the crawl size above it, and say what the crawl did not cover.

## Judgment

- Follow the [Google search notes](../create-seo-plan/references/platforms/google.md): [Search Console](../create-seo-plan/references/platforms/google.md#search-console) against estimates, the [credits](../create-seo-plan/references/platforms/google.md#credits) and the [handoff](../create-seo-plan/references/platforms/google.md#handoff).
- **Impact first.** Rank every issue by the traffic on the pages it touches, not by how many pages it touches. A noindex on the pricing page beats a thousand missing alt texts.
- `onpage_score` is DataForSEO's score, not Google's. Do not chase 100; chase the issues on pages that earn traffic.
- Many flags are cosmetic. `title_too_long` on a blog post, `no_image_alt` on decorative images and `has_misspelling` rarely move rankings. Mention them in one line, not one row each.
- The crawler follows links from the homepage, so a page nothing links to may never be crawled. A page the user expects that is missing from the crawl is itself a finding.
- There is no Lighthouse or field data in this crawl, and it does not load images, scripts or stylesheets, so `broken_resources` and render-blocking checks are not measured: a 0 there is not a pass. `high_loading_time` is the crawler's own timing; for Core Web Vitals the user checks PageSpeed Insights or Search Console's report.
- Results expire after 30 days. Crawl again after the fixes ship and compare the counts.
- Whether AI crawlers (GPTBot, PerplexityBot, ClaudeBot) can read the site is a different check with a different tool: [check-ai-crawler-access](../check-ai-crawler-access/SKILL.md). Point to it; do not fold it into this audit.

## Related skills

- Organic traffic or rankings fell: [diagnose-traffic-drop](../diagnose-traffic-drop/SKILL.md). A domain change, redesign or replatform: [create-migration-plan](../create-migration-plan/SKILL.md).
- Where the audit fits in a quarter's SEO work: [create-seo-plan](../create-seo-plan/SKILL.md). Keywords, quick wins and content: [find-seo-quick-wins](../find-seo-quick-wins/SKILL.md), [create-seo-content-plan](../create-seo-content-plan/SKILL.md).
- Whether AI crawlers can read the site, and llms.txt: [check-ai-crawler-access](../check-ai-crawler-access/SKILL.md). Which keywords show a Google AI overview and whether the site is cited in it: [check-ai-overviews](../check-ai-overviews/SKILL.md).
- Rankings and Search Console checked on a schedule: [track-rankings](../track-rankings/SKILL.md), [monitor-search-console](../monitor-search-console/SKILL.md).
- An audit for a prospect or a client: [prepare-client-pitch](../prepare-client-pitch/SKILL.md).