How to Run a Semrush Site Audit: Setup, Reports and Limits

What Site Audit checks, and why the limits matter more

Semrush's Site Audit crawls a website and lists what stands in the way of search engines and AI assistants: broken pages, slow loads, blocked bots, bad markup. The help center lists over 140 on-page and technical checks, from duplicate content and broken links to HTTPS, hreflang, AMP, crawlability and indexability ^[1]. A newer AI Search category checks readiness for tools such as ChatGPT search, Claude and Perplexity, and each issue comes with a How to fix note.

We read Semrush's help center from the US on October 12, 2026, and where it disagrees with the marketing pages, we say so. The short version is simple. The checks are thorough on every plan. What changes between plans is how many pages you can crawl and whether the crawler can render JavaScript, and those two limits decide whether Site Audit sees your whole site or only part of it.

Crawl budgets by plan

These figures come from the Site Audit help page. Starter inherits the Pro limits, Pro+ the Guru limits and Advanced the Business limits. Unused pages do not roll over; the budget resets on the first day of each month ^[1]. Your remaining budget sits under Subscription Info, in the Summary, as Pages to crawl, so check it before scheduling a new campaign.

What the limits mean in practice

A 20,000-page cap per audit covers most small and mid-size sites in one crawl. A store with 50,000 product URLs needs the Business tier, or several audits split by subfolder. Scheduling is where budgets quietly disappear. A 5,000-page site audited daily uses about 150,000 pages a month, which is beyond the Pro budget, by our arithmetic. Two other features sit behind higher tiers. JavaScript rendering needs Guru or Business, and AMP checks, with the option to crawl AMP pages first, come only with Business, although the Crawled Pages report still shows whether a page has an AMP version on any plan.

The free plan, the trial and Enterprise

The free plan crawls 100 pages a month, one audit at a time, which covers a small site's key pages. The 7-day trial unlocks paid limits but blocks exports ^[2]. Oddly, Semrush's own trial guide suggests exporting the issue list before the trial ends, which the trial rules do not allow. At the other end, Semrush sells Enterprise Site Intelligence above Advanced, which it describes as adding JavaScript rendering, crawling for millions of pages, compliance and QA checks, and accessibility management. No price is published. Our Semrush pricing guide covers the plans themselves.

Setting up your first audit

Project and scope

Your first Semrush project starts a Site Audit automatically for the domain you entered ^[3]. Later projects need a manual setup through a wizard of five steps, and only the first is required. Default settings are fine for a first pass. By default, Site Audit crawls the root domain with all subdomains and subfolders. To audit only a blog, enter the subdomain or subfolder and clear the Crawl all subdomains box. Narrow scopes save budget, but they also hide problems elsewhere, so run a full crawl at least once.

Page limit and crawl source

The Limit per audit field sets how many pages each crawl checks, and Semrush advises basing it on your plan and on how often you re-audit. Divide the monthly budget by the number of audits you plan, then leave a margin for single-page re-audits. There are four crawl sources: Website, the sitemap linked in robots.txt, a sitemap by URL, or a TXT or CSV file of URLs. A file is the cheapest way to check a specific set of pages. Only one sitemap URL can be used at a time, so sites with several sitemaps should upload a file instead.

Website versus sitemap

The Website source starts at the homepage and follows links breadth-first. That finds the pages closest to the homepage, which Semrush notes are not always the most important, while a sitemap crawl follows your own priorities. Run both once. The difference shows orphaned pages. Pages that appear in the sitemap but never in the website crawl are reachable only through the sitemap, so neither visitors nor crawlers following links will find them.

Schedule and user agent

Audits can run weekly on a chosen day, daily or once, and you can re-run at any time. Weekly suits most sites. Daily makes sense during a migration, when every deploy can break redirects, but switch back afterwards or the budget will run out. Three user agents are available: SiteAuditBot Desktop, SiteAuditBot Mobile and OpenAI-Search. Mobile is the default and mimics Google's mobile crawler, while Googlebot itself is no longer offered for new campaigns. Use OpenAI-Search to see your site as ChatGPT search sees it.

Crawler settings for difficult sites

Crawl delay and JavaScript

Minimum waits about one second between pages and is the default. The second option slows to one URL every two seconds, and the third respects the crawl-delay in robots.txt, which Semrush caps at 30 seconds. A crawl that is too fast can slow your site and produce false errors. JavaScript is the bigger decision. With rendering on, the crawler sees what visitors see; with it off, Site Audit reads only HTML, which is faster but less accurate on JavaScript-heavy pages. Rendering needs the Guru or Business tier ^[3]. On Pro or Starter, a site whose menus load by JavaScript may show only a handful of crawled pages.

Rules, parameters and staging sites

You can include or exclude subfolders by path, and the trailing slash matters: /shoes without it skips every URL that starts with /shoes, including /shoes-men. Check rules on a small crawl first. A wrong rule hides a whole section without warning. Listing parameters such as page, language or UTM tags makes Site Audit treat the variants as one URL. Without this, the crawler can check the same page twice and waste budget, which is why stores and large catalogs need it most.

Three options reach restricted pages. You can upload a verification file to bypass robots.txt and meta robots blocks, enter a username and password for protected areas, or use a Web Bot Auth signature. Some hosts, Shopify among them, block unknown bots by default, and a Web Bot Auth signature lets SiteAuditBot prove it is authorized. If an audit fails with an Auditing domain has failed error, the server is probably blocking the bot, and the crawl log can be sent to your developer.

Reading the Overview and the Site Health score

The Overview shows a Site Health score, errors and warnings, top issues, the AI Search Health score and thematic report widgets. It also breaks crawled pages into healthy, broken, with issues, redirected and blocked. Start with broken and blocked pages. They cost the most. Site Health itself is a percentage from 0 to 100, based on the errors and warnings found relative to the checks performed. Errors weigh more than warnings, and notices do not affect the score. A higher score means a lower density of problems, not a perfect site.

Errors are the most severe issues, warnings are medium severity and notices are informational. Each issue row shows the number of affected pages and a trend line, and clicking the trend line opens the Progress tab for that issue, which shows whether a fix held or the problem came back with the next deploy. Fix errors on pages that earn traffic first. A broken link on an old press release can wait. Connecting Google Analytics after setup helps here, because it surfaces issues on your most-viewed pages and fills the Pageviews column in Crawled Pages. Without it, every broken page looks equally urgent.

The thematic reports, one kind of fix each

The Overview leads to eight thematic reports: robots.txt, Crawlability, HTTPS, International SEO, Performance, Internal Linking, Markups and Core Web Vitals. Each scores and groups related issues, and they are faster to act on than the full issues list because each one maps to one kind of fix. A developer can take Performance while a content editor takes Internal Linking, without either of them reading the other's list.

Crawlability and internal linking

The Crawlability report shows crawl depth, HTTP status codes and a sitemap versus crawled pages comparison. Semrush recommends keeping important content within three clicks of the homepage, and a big gap between sitemap and crawled counts usually means weak internal linking or technical blocks. The Internal Linking report flags pages with high click depth, orphaned pages with no internal links, broken links and the pages passing the most Internal LinkRank, a measure of how reachable a page is from incoming links and crawl depth. Pages with one incoming link are the easiest wins. Add two more.

Performance and Core Web Vitals

The Performance report checks HTML size, redirect chains, load speed, compression, caching and the size and number of JavaScript and CSS files. It groups pages by HTML load time: under 0.5 seconds, 0.5 to 1, 1 to 3, and over 3. Semrush recommends under one second, though HTML load time ignores images and scripts. The Core Web Vitals report measures Largest Contentful Paint, Cumulative Layout Shift and Total Blocking Time with Google Lighthouse, on a set of 10 pages chosen manually or automatically. Total Blocking Time stands in for First Input Delay, which cannot be measured in a lab. Ten pages is a sample, not a site. Choose them.

The lab conditions explain why the numbers rarely match Search Console. Mobile tests emulate a 360 by 640 screen on slow 4G with four-times CPU throttling, and desktop uses 1350 by 940 on 10 Mbps. Semrush's servers are in the US, which can add delay for sites elsewhere, and Semrush itself recommends Google PageSpeed Insights for comparison. Expect the numbers to differ.

Markups and International SEO

The Markups report grades structured data by the share of valid items, lists markup types and shows each item found. Site Audit checks nearly every structured data type in Google's vocabulary, except beta ones, and Semrush's trial guide notes that invalid markup can cost review stars and other rich results. The International SEO report checks hreflang tags for wrong codes and missing language versions, but only in the HTML, not in sitemaps or HTTP headers. If your hreflang lives in the sitemap, this report will show nothing useful. Semrush also says its missing-tag data is informative only.

AI Search Health and the eight AI bots

AI Search Health is a separate score for AI readiness ^[4]. It counts AI-related issues, whether AI bots are blocked in robots.txt, and missing elements such as schema, crawlable navigation, internal links and llms.txt. Filter the Issues tab by AI Search to see the specific problems. Site Audit checks access for ChatGPT-User, OAI-SearchBot, Googlebot, Google-Extended, Perplexity-User, PerplexityBot, Claude-User and Claude-SearchBot. Semrush groups them into on-demand fetchers, search crawlers, training bots and traditional search bots, and it says blocking training bots such as GPTBot or Google-Extended does not affect search visibility.

Working through the fixes

Crawled Pages and single-page re-audits

The Crawled Pages report lists every crawled page with more than 20 optional columns, including Internal LinkRank, blocked AI bots, status code, crawl depth, pageviews, load time, markup and canonicals. Combine filters to find pages with issues and an Internal LinkRank above 70, then sort by pageviews, which need a connected Google Analytics account. The Reaudit column re-crawls a single page without spending a full audit. Use it after each fix. It confirms the change and keeps the budget for the next scheduled crawl.

Hiding issues, honestly

Issues can be hidden entirely, hidden on specific pages, or excluded in bulk from future checks, and hidden issues no longer affect Site Health. That is useful for deliberate choices, such as nofollow internal links. It is also an easy way to flatter a score. Agencies should list what they hid in client reports, so the client can see that a rising score reflects real fixes rather than a shorter list of checks.

Handing work over and showing progress

Each issue can be sent to Trello as a card, or shared with Viewer or Editor access, and CSV exports cover other task tools. Send issues in batches by template, not one URL at a time. One fix to a template often clears hundreds of rows. When the work is done, Compare Crawls sets any two audits side by side and counts fixed and new issues, while the Progress tab charts any issue or metric over time. Both export to PDF or Excel. This is the report that shows a client what the work achieved, because it shows which problems disappeared and which new ones arrived. Around it, a simple routine on a weekly scheduled audit keeps the work manageable:

• Each week: check broken and blocked pages and new errors in Compare Crawls.
• Each month: review the Internal Linking and Performance reports.
• Each quarter: run a full crawl with all subdomains.
• After every fix: re-audit the page, not the whole site.

When the crawl goes wrong

If only four to six pages appear, the bot probably found no internal links on the homepage, often because they sit inside JavaScript. The fix on Pro or Starter is to crawl from a sitemap or a URL file. Size can also block a crawl, since homepages over 4 MB and other pages over 2 MB cannot be parsed. A different problem is the audit that seems stuck. The last stage checks external links and images, so an audit can pause after the page counter stops.

You can stop a partial crawl and save it as the latest audit, which preserves the rest of the budget, or discard it, which also saves budget but leaves no results. If it stays stuck for more than 24 hours, Semrush asks you to contact support. Site Audit keeps data for up to one year after the last crawl, and a campaign not re-run within 12 months is deleted with its settings, after a notice. To delete a single Site Audit campaign, you must delete the whole project and every tracking campaign in it. Export before deleting.

The bottom line

Semrush's Site Audit is thorough for a general SEO suite: 140+ checks, AI bot access and clear priorities. Its limits are the real decision. Pro and Starter cap each audit at 20,000 pages and skip JavaScript rendering, which matters for modern front ends, so large or script-heavy sites should price Pro+ or Advanced before relying on it. For a small business site with a few hundred pages and plain HTML menus, though, the Pro limits are generous, and even the free plan's 100 pages give an honest first picture of what needs fixing. Our Semrush review covers the rest of the suite.