How to Read an External Site Audit: What a Source-Blind Crawler Sees
The sections of an audit report produced by crawling only over HTTP, what each finding means, which ones are urgent, and what such a report cannot tell you. For checking your agency's or an AI's report.
Short answer: An external audit that never reads source code shows the site as search engines and AI crawlers see it: which pages exist, which are indexable, whether canonical and hreflang are consistent, whether the sitemap matches the real pages, which pages receive no internal link at all, whether bots are blocked. It is read in this order: scope, indexability, link architecture, schema and sitemap, then speed. There are things such a report cannot say, and a good one writes them down as "not measured".
This article was written so that you can check an audit report that lands on your desk, whether ours, another agency's or one produced by an AI. Our own crawler works on exactly this principle: it only looks from outside, it is the first step of our service for existing websites, and it repeats the same measurement at the end of the work.
0. Scope: what did the report crawl?
Every report should open with three numbers: how many URLs were found, how many were crawled, whether the crawl hit a limit. If 50 pages of a 200-page site were crawled, a "no orphan pages" finding is meaningless; the crawl stopped halfway. Were subdomains (blog., shop.) counted as separate sites or included; which user agent was used. If those lines are missing, do not trust the report.
1. Indexability
Findings. noindex pages, paths closed by robots.txt, internal links returning 4xx and 5xx, redirect chains.
Meaning. A search engine cannot rank a page it cannot index. For a site that is red here, none of the findings above it are a priority.
Common trap. If the homepage redirects with 301/308 to another version (http → https, apex → www) and the crawler does not follow it, the whole site looks orphaned. The report should state the homepage's final address.
2. Canonical and hreflang
Findings. Canonical not pointing at itself, canonical pointing at the homepage, missing or contradictory hreflang on a multilingual site, language versions not matched one to one.
Meaning. Canonical tells the search engine "this is the main version"; if it is wrong, the page is folded into another page in the search engine's eyes. If hreflang is missing, language versions compete with each other.
How to check. Open a page the report gives as an example and look for the rel="canonical" and hreflang lines in its source; that is the strength of an external audit, anyone can repeat it.
3. Link architecture
Findings. Orphan pages (no internal link at all), money pages linked only from the menu or the footer, depth (how many clicks from the homepage), links to the same page always with the same text, leakage between languages.
Meaning. Internal linking is the strongest signal telling a search engine which page matters. If a service page is orphaned or linked only from the footer, the guide content is not carrying it.
Scale. A good report gives this as a list, not a score: which page is orphaned, how many links each money page gets, from where.
4. Schema and sitemap
Findings. Which structured-data type sits on which page, whether it parses, whether the Organization node is single; whether the URLs in the sitemap match the crawled URLs, whether the sitemap contains a foreign host or redirected URLs.
Meaning. Schema is the measurable part of the understood layer. The sitemap is the "crawl these" list for the search engine; pages that exist but are not in it, and pages in it that return 404, are two separate problems.
5. Speed and weight basics
Findings. Page weight, image format and size, blocking scripts; an external crawler can measure these. Real-user speed data (Core Web Vitals) comes not from an external crawl but from Google's field data and Search Console.
Meaning. An external crawl says "this image is 3 MB"; saying "the site is slow" needs field data. Be careful if a report mixes the two.
6. Bot access
Findings. The response code the same page gives to different user agents (browser, Googlebot, bingbot, AI crawlers).
Meaning. A host's bot filter sometimes cuts Bing or AI crawlers off with 403; the site is on Google but appears absent from Bing and from AI answers. Saying "we are not on Bing" or "AI does not see us" without measuring this is a wrong diagnosis.
What the report cannot say
- Rankings, impressions, clicks: Search Console.
- Visitor behaviour and conversion: Analytics and contact measurement.
- Source code and server settings: admin panel and developer.
- Why: an external crawl says "canonical is wrong", not "why it is wrong".
A good report states these limits; a report with no "not measured" line is not complete, it is over-confident.
Test the report in your hands with three questions
- Is there a scope line: how many URLs, how many crawled, did it hit a limit?
- Does every finding carry an example URL that you can open and see for yourself?
- Does the report say what it could not measure?
If you cannot answer "yes" to all three, the report is a sales text, not a diagnosis.
How does your site look from outside?
The same measurement sits inside the Check-up report, item by item with example URLs. If the findings say "fix", how the layers are added without a rebuild is on the SEO, GEO and AEO for your existing website page.
01What is an external site audit?
A crawl done without looking at the site's source code or admin panel, using only public HTTP responses. It shows what search engines and AI crawlers see; unlike an internal report, it can be repeated by anyone regardless of who produced it.
02What should be read first in an audit report?
Indexability: how many pages were crawled, how many are indexable, where canonical points, whether there are orphan pages. If this section is bad, schema, speed and content findings come second.
03Does an audit report show rankings or traffic?
No. An external audit shows the site's technical state; rankings, impressions and clicks come from Search Console, visitor behaviour from Analytics. The two are read together.
Related service SEO, GEO and AEO for Your Existing Website →