Give it a list of domains and get back one scored row per website: a 0 to 100 technical SEO score, a grade, and a plain list of what is wrong. Each row covers robots.txt, sitemap health, status codes, redirect chains, soft 404s, structured data, title and meta tags, security headers, AI crawler policy and feeds. No API key, no proxy, no browser, and a domain that cannot be audited is never charged.

It is built for checking many sites quickly, not for crawling one site deeply. Twenty to thirty requests per domain, a few seconds each, one row you can sort and filter.

Who uses it

• SEO agencies and freelancers qualifying prospects: run 200 lead domains, sort by score, and open the conversation with the three issues each site actually has.
• In-house SEO and web teams keeping a weekly eye on every property they own, including the microsites nobody remembers.
• Sales and lead-generation teams selling web, hosting or SEO services, who want a reason to call that is specific to the prospect's own site.
• Buyers of websites and domains doing a first technical pass before paying for a deeper audit.
• AI agents and automations that need a structured answer to "is this site technically healthy?" in one call.

What gets checked

Check	What it answers	Columns
Status codes	Does the homepage load, and do a sample of the site's pages load	homepageStatus, pagesSampled, pagesOk, pagesBroken, brokenPages
Redirect chains	How many hops before the homepage loads, and does http end up on https	redirectHops, redirectChain, httpsAvailable, httpRedirectsToHttps
robots.txt	Is there a real one, and how restrictive is it	robotsTxtFound, robotsTxtUrl, robotsDisallowRules
Sitemap health	Is there a sitemap, is it listed in robots.txt, how many URLs, and do the URLs in it actually work	sitemapFound, sitemapUrl, sitemapInRobotsTxt, sitemapIsIndex, sitemapUrlCount, pagesRedirected
Soft 404s	Does a made-up address return a real 404, and do sampled pages answer 200 while looking dead	missingPageStatus, soft404Pages, soft404Rate
On-page tags	Title, meta description, canonical, H1 count, noindex, language, Open Graph	title, titleLength, metaDescription, descriptionLength, canonical, h1Count, homepageNoindex, lang, hasOpenGraph
Structured data	Does the homepage carry valid JSON-LD, and of which schema.org types	structuredDataFound, structuredDataTypes
Security headers	How many of the eight standard headers are sent	securityHeadersScore, securityHeadersGrade, securityHeadersMissing
AI crawler policy	Which AI crawlers robots.txt blocks by name, and is there an llms.txt	aiCrawlersBlocked, aiCrawlersNamed, hasLlmsTxt
Feeds	Is there a working RSS, Atom or JSON feed	feedFound, feedUrls

Pages are sampled evenly across the sitemap, so a 1,600-URL sitemap is not judged by its first ten lines. When a site has no sitemap, the sample is taken from the homepage's own links instead and sampleSource says so.

How the score is computed

The score is a fixed, published rule, not a black box. scoreBreakdown on every row shows the points earned and the points possible for each line, so you can re-add it by hand.

Check	Points	Full marks when
Homepage	10	The homepage ends on a 2xx status
HTTPS	10	https works (5) and http redirects to it (5)
Redirect chain	5	At most one hop before the homepage loads; two hops earn 3
robots.txt	10	A real robots.txt is served
Sitemap	10	A sitemap with at least one URL is found
Sampled pages	15	Every sampled page ends on a 2xx status (proportional otherwise)
Soft-404 pages	5	No sampled page looks like a soft 404 (proportional otherwise)
Missing page	5	A made-up address returns 404 or 410
On-page tags	10	Title (3), meta description (3), canonical (2), exactly one H1 (2); a noindex homepage earns 0
Structured data	10	The homepage has at least one valid JSON-LD block
Security headers	10	All eight headers present (proportional otherwise)

A check that could not be made is left out of the total instead of being scored as a failure. A site with no pages to sample is scored out of 80, then scaled to 100. Grades: A from 90, B from 75, C from 60, D from 40, F below.

AI crawler policy and feeds are reported but never scored: blocking GPTBot is a choice, and not every site needs a feed.

The issues column carries codes from a fixed list: HOMEPAGE_NOT_200, NO_HTTPS, HTTP_NOT_REDIRECTED_TO_HTTPS, LONG_REDIRECT_CHAIN, NO_ROBOTS_TXT, NO_SITEMAP, SITEMAP_NOT_IN_ROBOTS_TXT, BROKEN_PAGES, SITEMAP_URLS_REDIRECT, SOFT_404_PAGES, MISSING_PAGE_NOT_404, MISSING_TITLE, TITLE_TOO_LONG, MISSING_META_DESCRIPTION, MISSING_CANONICAL, CANONICAL_POINTS_ELSEWHERE, MISSING_H1, MULTIPLE_H1, NO_VIEWPORT_META, HOMEPAGE_NOINDEX, NO_STRUCTURED_DATA, INVALID_JSON_LD, NO_HSTS, WEAK_SECURITY_HEADERS.

How to use Technical SEO Audit

1. In the Apify Console. Open the actor page and click Start — the domains field is already pre-filled with a working example. Results land in the run's dataset as soon as each item is found.
2. Via the API. Call it directly with a POST request — no Console needed once you have an API token:

   curl "https://api.apify.com/v2/acts/accountable_eel~technical-seo-audit-composite/run-sync-get-dataset-items?token=<YOUR_TOKEN>" \
     -X POST \
     -H "Content-Type: application/json" \
     -d '{"domains":["example.com","python.org","nodejs.org"]}'
   

3. On a schedule. Save this actor as an Apify Task with the input you want, then add a Schedule (hourly, daily, weekly) so it runs on its own — no server of your own required.

sampleSize sets how many pages of each site are opened (10 by default, 25 at most, 0 to skip page sampling). It changes how thorough the row is, never the price.

Input

{
  "domains": [
    "example.com",
    "python.org",
    "nodejs.org"
  ]
}

One domain per line, such as example.com. A full address works too: only the domain is used, and the audit always starts from the homepage. Each domain comes back as one scored row. Accepted formats: example.com, www.python.org, https://nodejs.org/en/about.

A full address such as https://nodejs.org/en/about is accepted, but only the domain is used: the audit always starts from the homepage. python.org and www.python.org are both fine, and the domain column tells you which host the site actually sent visitors to.

Output

One row per domain.

query	found	status	domain	score	grade	issues	issueCount	scoreBreakdown	homepageStatus	homepageResponseMs	httpsAvailable	httpRedirectsToHttps	redirectHops	redirectChain	robotsTxtFound	robotsTxtUrl	robotsDisallowRules	sitemapFound	sitemapUrl	sitemapInRobotsTxt	sitemapIsIndex	sitemapUrlCount	sampleSource	pagesSampled	pagesOk	pagesRedirected	pagesBroken	brokenPages	soft404Pages	soft404Rate	missingPageStatus	title	titleLength	metaDescription	descriptionLength	canonical	h1Count	homepageNoindex	lang	hasOpenGraph	structuredDataFound	structuredDataTypes	securityHeadersScore	securityHeadersGrade	securityHeadersMissing	aiCrawlersBlocked	aiCrawlersNamed	hasLlmsTxt	feedFound	feedUrls	sourceUrl	requestsMade	scrapedAt
example.com	true	OK		<score (0-100)>						<homepage response time (ms)>	<https works?>	<http redirects to https?>			<robots.txt found?>	<robots.txt address>		<sitemap found?>		<sitemap listed in robots.txt?>	<sitemap is an index?>													<meta description>				<homepage set to noindex?>		<has open graph tags?>	<structured data found?>		<security headers present (of 8)>					<has llms.txt?>	<feed found?>				1970-01-01T00:00:00.000Z

Three real rows from one run on 2026-10-11, trimmed to the columns that tell the story:

query	domain	score	grade	issues	redirectHops	sitemapUrlCount	sampleSource	pagesSampled	pagesBroken	missingPageStatus	structuredDataTypes	securityHeadersGrade	hasLlmsTxt	feedFound
example.com	example.com	35	F	HTTP_NOT_REDIRECTED_TO_HTTPS, NO_ROBOTS_TXT, NO_SITEMAP, MISSING_META_DESCRIPTION, MISSING_CANONICAL, MISSING_H1, NO_STRUCTURED_DATA, NO_HSTS, WEAK_SECURITY_HEADERS	0					404		F	false	false
python.org	www.python.org	79	B	NO_SITEMAP, MISSING_CANONICAL, MULTIPLE_H1, WEAK_SECURITY_HEADERS	1		homepage-links	10	0	404	WebSite	D	false	true
nodejs.org	nodejs.org	77	B	BROKEN_PAGES, NO_STRUCTURED_DATA, WEAK_SECURITY_HEADERS	1	1653	sitemap	10	4	404		D	true	true

The third row is the kind of thing this actor exists to find. The site's own sitemap lists 1,653 URLs, and four of the ten sampled ones return 404: the sitemap carries release addresses with a dot missing (/en/blog/release/v08.20 where the page lives at v0.8.20). brokenPages lists each one with its status code.

An empty cell is a real "not known", never a hidden zero: sitemapUrlCount is empty when there is no sitemap, and the sampled-page columns are empty for a one-page site with nothing to sample. A 0 is always a measured zero.

A domain that could not be audited comes back as a row with found: false, a status and a message that says why, and is never charged:

status	Meaning
NOT_FOUND	The domain does not resolve, or did not answer over https or http
BLOCKED	The homepage answered with a bot check or a refusal instead of the page
ROBOTS_DISALLOWED	robots.txt forbids the whole site to every crawler, so no page was fetched
BAD_INPUT	The line is not a public website domain

Example runs

Three inputs that return real rows with no key or account of your own.

1. The prefilled example: three very different sites. A one-page placeholder, a large site with no sitemap, and a large site with a 1,653-URL sitemap. This is the run the rows above come from:

{
  "domains": ["example.com", "python.org", "nodejs.org"]
}

It comes back as F (35), B (79) and B (77). The middle row shows the fallback: no sitemap, so ten pages are taken from the homepage's own links and sampleSource reads homepage-links.

2. A deeper look at one site. Raise the sample to the maximum to check 25 pages spread across the sitemap:

{
  "domains": ["nodejs.org"],
  "sampleSize": 25
}

3. A fast pass with only the columns a prospect list needs. Page sampling off, four columns kept:

{
  "domains": ["python.org", "example.com"],
  "sampleSize": 0,
  "columns": ["domain", "score", "grade", "issues"]
}

With sampling off the sampled-page columns stay empty and those checks are left out of the score.

How a domain gets billed

One event, domain-audited, charged once per domain that was actually audited.

• The price is per domain, whatever the sample size and however many of the checks found something.
• A domain that does not resolve, answers with a bot check, forbids crawling in robots.txt, or is not a domain at all comes back found: false and is not charged.
• A homepage that answers with a server error is audited and charged: "your homepage returns 500" is the finding.
• Rows removed by your own keyword filters, or past "Stop after this many results", are not charged.

Pricing

Pay-per-event. A flat per-run fee covers session/proxy warmup; each independent source is billed under its own event, only when that source actually returns data — see .actor/pay_per_event.json for exact prices. A miss is never charged.

$18 per 1,000 domains on the free plan, less on paid Apify plans. There is no per-page or per-check charge on top.

Limits worth knowing before you buy

• It samples, it does not crawl. One homepage plus up to 25 sampled pages per domain. A crawler that walks 500 pages of one site will find more on that site; this finds the site-wide faults fast across many sites.
• No Lighthouse, no Core Web Vitals, no backlinks, no keyword rankings. Nothing here runs a browser. homepageResponseMs is a server response time from a cloud server, not a page-speed score.
• On-page tags and structured data are read from the homepage only, as the server sends it. Tags that a site adds with JavaScript after loading are not seen.
• For a sitemap index, the first child sitemap is read. sitemapUrlCount counts that one file.
• Sites behind a bot wall are not audited. No proxy is used and no challenge is solved; such a domain is returned as BLOCKED for free.
• robots.txt is respected. Sampled pages that the site disallows for every crawler are skipped, and a site that disallows everything is not fetched at all.
• The soft-404 check is conservative. It flags a page that redirects to the homepage, is titled as a not-found page, or matches what the site serves for a made-up address. It is tuned to avoid false alarms, so it will miss some.

Frequently asked questions

How long does a run take? A few seconds per domain. Three domains finished in five seconds in the example run; domains are audited three at a time by default.

Do I need a proxy or an API key? No. The actor makes plain requests from Apify's own servers and identifies itself honestly in its user agent.

Why is my score not out of the same total as another site's? Checks that could not be made are left out. scoreBreakdown shows exactly which lines counted.

Can I run it every week and compare? Yes. Save it as a task and add a schedule, then compare score and issues between runs in your own sheet or workflow. The actor itself keeps no memory between runs.

Why does domain differ from what I typed? It is the host the site redirected to. The address you typed stays in query.

Is blocking AI crawlers counted against the site? No. aiCrawlersBlocked is reported so you can see the policy, and it never changes the score.

Use it from Clay, n8n, Make, or an AI agent

This actor runs synchronously over plain HTTP — call it directly from a script, a workflow tool, or an AI agent, no Apify Console needed once you have an API token.

curl "https://api.apify.com/v2/acts/accountable_eel~technical-seo-audit-composite/run-sync-get-dataset-items?token=<YOUR_TOKEN>" \
  -X POST \
  -H "Content-Type: application/json" \
  -d '{"domains":["example.com","python.org","nodejs.org"]}'

n8n. Add an HTTP Request node: Method POST, URL https://api.apify.com/v2/acts/accountable_eel~technical-seo-audit-composite/run-sync-get-dataset-items?token=<YOUR_TOKEN>, Body Content Type JSON, JSON Body {"domains":["example.com","python.org","nodejs.org"]} (swap in an expression from an earlier node for a real value).

Clay. Add an "HTTP API" column: Method POST, URL https://api.apify.com/v2/acts/accountable_eel~technical-seo-audit-composite/run-sync-get-dataset-items?token=<YOUR_TOKEN>, Body {"domains":["{{domain}}"]}, mapping the row's domain into the domains array.

MCP. In Claude, Cursor, or any MCP client with the Apify MCP server, ask for "Technical SEO audit tool: one scored row per domain" — the agent will find and run this actor.