Site Audit
A crawl of your whole site, read page by page in the background, then what to fix first – broken links inside the site and to other sites, redirect chains, pages Google can't index, missing and duplicate titles, slow and heavy pages, images without alt text – with Lighthouse on the pages that matter most.
Open it from Studios. Three tiles open the same page: Site Audit for the whole report, Broken Link Checker for the broken links, and Page Speed for measuring any page with Lighthouse.
Every number in an audit is counted from the crawl, measured by Lighthouse or read from the links other sites have to yours. None is estimated, and there is no made-up health score.
Start an audit
The site. The brand's website is filled in from its profile. Change it to audit another site the brand runs.
How many pages. 100, 250, 500, 1,000, 2,500 or 5,000 – 500 unless you pick another. The crawl starts at the home page and follows the links it finds until it has read every page it can reach or reaches the limit.
JavaScript rendering (off unless you turn it on). Some sites build their pages in the browser, so their links and text appear only after scripts run. With rendering on, each page is opened the way a browser opens it. It costs about four times as much per page and takes longer, so turn it on only when an audit without it finds far fewer pages than the site has.
The price shows before you start, as up to a number of credits – see What it costs. Nothing on your site is changed: the crawl only reads it, and it waits about a second between pages so it adds little load to your server.
One crawl per site at a time. Starting an audit of a site that is already being crawled for the brand – from this page, from chat or by your AI team – opens that crawl instead. Nothing is charged twice.
While it crawls
The crawl runs on our servers, so you can close the page. The page shows how far it has got – the pages read against the limit – then the steps after the crawl: checking links to other sites, measuring the key pages with Lighthouse and writing the report. When a crawl you or chat started finishes, the bell rings with what it found first, and the audit opens from Your audits.
With its one-second pause between pages, a crawl reads a page about every 1.7 seconds: 500 pages take about 15 minutes and 5,000 about two and a quarter hours, plus a few minutes for the steps after the crawl. A slow server takes longer, and so does JavaScript rendering.
Stop and keep what's crawled ends a crawl while it runs. The pages read so far become the audit, and you pay for those pages only. A crawl that runs far past the time its page limit should take – usually a server that answers very slowly – is stopped the same way.
The report
The report opens with its figures – the kinds of error and warning, which together are what to fix first, broken links, pages Google can't index and the home page's speed – and says whether the crawl read every page it found or stopped at its limit. Your audits shows each audit's count to fix first the same way.
Fix first comes next: up to five findings, the most serious first, then the ones on the most pages. Each says what it is in plain words, how to fix it, and the pages it was found on – or Whole site for a finding about the site itself. Everything else it found lists the rest the same way. Each finding is weighed as one to fix first, one worth fixing, or one good to know.
What the audit checks:
| Area | What it checks |
|---|---|
| Links and redirects | Pages that answer with an error (404, 410, 500 and the like), links inside the site to them, links to other sites that are dead, redirect loops, redirect chains of two hops or more, links that go through a redirect instead of straight to the final page, meta refresh redirects |
| Indexing | Missing pages that don't answer 404 (see below), canonical tags that point at a broken page, a redirect or another canonical, or in a circle, pages kept out of Google and why, no sitemap, no robots.txt, www and the other versions of the address not leading to one with a permanent redirect |
| Titles and content | Pages without a title, titles shared by several pages, titles longer than Google shows or very short, pages without a meta description, descriptions shared by several pages, pages without an H1, pages with the same content, pages with very little text, placeholder text left on pages, outdated HTML tags, no doctype, no character set, no favicon |
| Speed | Slow pages, a slow server response, heavy pages, render-blocking files |
| Images | Broken images, images without alt text |
| Security | No valid HTTPS certificate or one that expires within 21 days, pages served without HTTPS, secure pages linking to insecure ones |
Checks that no longer matter to Google, such as meta keywords, are left out.
Below Fix first, the report goes through the site by topic:
- Broken links – see Broken Link Checker.
- Redirects – each chain with every hop and its status code, loops marked, and how many pages link to the first address. Point those links straight at the final page.
- Indexing and site-wide checks – each page Google can't index, with its reason: blocked by robots.txt, a noindex tag, an HTTP header, a canonical tag that names another page, or too many redirects. A page listed here on purpose (a login page, a thank-you page) needs nothing. Then the whole site: HTTPS and the certificate, whether the http:// address sends visitors on to https://, whether the address variants lead to one, what a missing page answers, the sitemap, robots.txt and HTTP/2.
- Titles, descriptions and content – titles and descriptions shared by more than one page, grouped with the pages that share them, and the pages with a missing or overlong title or description or no H1.
- Speed – Lighthouse on mobile for the home page and the two pages the site links to most: the performance, accessibility, best practices and SEO scores (0–100), Largest Contentful Paint, Cumulative Layout Shift, Total Blocking Time, First Contentful Paint and Speed Index, and up to five opportunities with the time or bytes each would save. It is included in the audit's price. The crawl times how long each page takes to arrive; how fast a page shows on a phone comes from Lighthouse.
- Images – broken images and images without alt text, with their pages.
- All pages – the pages read, up to 2,000, the ones with the most findings first: what each answered, its words, how many clicks it is from the home page and its findings. Find a page by its path or title.
Broken Link Checker
The broken links from the brand's latest audit, inside the site and to other sites. Each shows the address, what it answered (a status code, or that the site doesn't exist or didn't answer), whether it is a link, an image, a script, a stylesheet, a canonical tag or a redirect, the words it was linked with, and the pages it was found on. Fix the links on those pages, or send the dead address on to the page that replaced it with a 301 redirect. With no audit yet, start one here.
When missing pages answer 200
A well-behaved site answers an address that doesn't exist with 404 Not Found. Some sites answer it with 200 OK – the code for a real page – and show the home page or an empty page instead. It is common on sites built as a single-page app.
On such a site, Google can't tell a dead page from a real one, and no crawler – this one included – can find a broken link inside the site, because every address answers as if it works. The audit makes it a finding of its own, Missing pages don't answer 404, for the whole site, and the broken links say that links inside the site may be missing from them. Have your developer make missing addresses answer 404, then run the audit again.
Links to other sites that couldn't be checked
Links to other sites are checked from our servers once the crawl ends. Some sites turn automated checks away: they answer forbidden (403) or too many requests (429), ask for a login (401), answer with a code of their own, or don't answer in time. Those links are counted apart, as couldn't check – never as broken and never as working – and the report says how many there were. An audit checks up to 400 addresses on other sites; when a site links out to more, it says how many were checked out of how many.
Links the crawl didn't open
A link inside the site to an address the crawl did not open is not checked, and the report counts those links. When the crawl stopped at its page limit, the report says so; if the count is high, run the audit again with a higher limit.
Links from other sites to your dead pages
Other sites may still link to pages of yours that no longer work – visitors and ranking sent to a dead end. Under Links from other sites to dead pages, Find them reads the links from other sites we know about that land on a page of yours that no longer answers: the page they are on, the words they use, the dead page they point at and what it answers. It costs 2 credits each time. Send each dead page on to its closest live page with a 301 redirect, and those links count for your site again.
Page Speed
The Speed tab measures any page with Lighthouse – up to 3 addresses at once, on mobile or desktop, in about 20 seconds. Each page gets its four scores, the five measurements and up to five opportunities, as in the audit. It costs 1 credit an address; an address Lighthouse can't measure is refunded.
Your audits, the Library and exports
Your audits lists the brand's audits, newest first. Open one to read it again – opening an audit costs 0 credits. Everyone with access to the brand can open its audits.
Each finished audit is also saved to the Library as Site audit – your domain, and opens here.
CSV downloads the broken links, the pages or the findings, for a spreadsheet or your developer's ticket list. PDF, Word and Slides turn the audit into a report to share.
In chat
Ask chat to "audit our site", "find broken links on example.com" or "why aren't our pages indexed" and it runs the same crawl, at the same price – on the brand's website when you don't name one. Chat needs a brand.
- While it crawls, a card under the reply follows the crawl and fills in when it finishes – the pages read, what to fix and the broken links, the first three findings, and Open the full audit. The bell rings too. If the crawl is still running after three hours, the card stops following it and links to Site Audit.
- A recent audit is reused. When the brand's site was audited in the last 7 days, chat answers from that audit at 0 credits and says its date. Say "re-run" for a fresh crawl.
- Ask about it any time. "What did the audit find?" or "which redirects should we fix?" reads the latest audit at 0 credits.
- Broken links gives the broken links from the latest audit (starting one when there is none), with the links from other sites that land on your dead pages (2 credits).
- Page speed measures up to 3 addresses with Lighthouse, on mobile unless you ask for desktop, in about 20 seconds – 1 credit an address.
- Check a fix. Once your developer has fixed something, ask chat to check it again: it re-reads up to 20 pages and the broken links behind the findings, and says for each whether it is fixed, still there or changed – 1 credit.
With the SEO Strategist
The SEO Strategist reads the brand's latest audit as part of its research – the problems by kind, the broken links and the page speed. When the latest audit is more than 14 days old, or there is none, it starts a new one – at most 1,000 pages, paid from your credits like any other audit – and carries on without waiting, with the older audit (dated) or without one. Its next batch of ideas reads the new one. A crawl your AI team starts rings no bell, and is listed under Your audits and in the Library like any other.
A technical fix it proposes names the findings it fixes – the broken links, redirects or missing titles from the audit – and a goal made from it ends by checking those pages and links again (1 credit), so the goal shows what is fixed and what is still there.
What it costs
A crawl is priced on the pages it reads. The price before you start is the most it can cost – every page up to the limit, and Lighthouse on three key pages:
| Pages | Up to | With JavaScript rendering, up to |
|---|---|---|
| 100 | 3 credits | 4 credits |
| 250 | 3 credits | 10 credits |
| 500 | 5 credits | 19 credits |
| 1,000 | 10 credits | 37 credits |
| 2,500 | 23 credits | 91 credits |
| 5,000 | 46 credits | 181 credits |
When the crawl ends it is settled on the pages it read, at least 3 credits, and the rest comes back to your credits – a 500-page audit of a 40-page site costs 3. A crawl that fails is refunded in full. Your credit history shows the reserve and what came back as separate lines.
| Also | Price |
|---|---|
| Opening or asking about an audit – in the page, the Library or chat | 0 credits |
| Page speed | 1 credit an address |
| Links from other sites to your dead pages | 2 credits |
| Checking a fix again | 1 credit |
Not covered yet
- The API. Site Audit, broken links and page speed are not available over the API yet.
- Scheduled audits. An audit runs when you, chat or your AI team start it. Run it again after a round of fixes; each audit stands on its own.
- Rendering by default. JavaScript rendering is a switch, off unless you turn it on.
- hreflang and structured data. Language alternates and structured data are not checked across the site. SEO Meta & Schema checks one page's structured data.
- Pages behind a login. The crawl reads the site as a visitor who is not signed in.
- Your AI team's spending limits. When the team starts a crawl, the part of its price that comes back is returned to your credits but is not taken off the team's daily and weekly spending limits.