What a Website Audit Is, and What It Actually Finds

Artyom Sklyarov··9 min read

A website audit is a structured check of whether search engines can crawl and understand a site, whether people can load and use it, and whether anything on it is quietly costing traffic or trust. A good one covers technical health, on-page basics, content, speed, structured data, readiness for AI search and images, and says which problems matter. Most end in a score out of 100, and the score is the least useful part. We have audited our own site, suur.io, five times since July. The score went 69, 64, 62, 69, 71. Over the same weeks we found a homepage with 33 seconds of blocking time on mobile, 92 tag pages with one post each, and a site where every URL passed every hygiene check while it ranked for eight keywords. Below: what an audit covers, what those five found, what a score can and cannot tell you, and what an audit costs.

The historical evidence extract publishes the retained August and September report figures and their limitations. It is an internal-report extract, not the raw Search Console exports or an independent reproduction.

What an audit is, and what it is not

An audit is a diagnosis. It reads the site the way Google and a visitor would, compares what it finds against known failure modes, and ranks the problems by what they cost. The output is a list of fixes in order, not a grade.

It is not a crawl report. A crawler such as Screaming Frog or Ahrefs Site Audit will hand you hundreds of warnings, most of them harmless. The audit is the reading of that report: which three of the four hundred rows are worth a developer's week.

It is also not a strategy. An audit tells you what is broken or missing on the site you have. It does not tell you which searches are worth winning or what to publish; that is the next document, and a good audit ends by pointing at it.

What a website audit covers

Our audits score seven areas, weighted by how much each tends to matter. The weights are the audit's opinion, not a law, but most serious audits cover the same ground:

AreaWeightWhat it checks
Technical SEO22%Can Google crawl and index every page you want found, and none you do not? Status codes, redirects, canonicals, robots rules, the sitemap, HTTPS
Content quality23%Is each page useful, original and credibly written? Thin and duplicate pages, author and evidence signals
On-page SEO20%Titles, descriptions, headings, internal links, social share cards
Structured data10%Schema markup that is valid and describes the page correctly
Performance10%Core Web Vitals: how fast the main content appears, how quickly the page responds, whether it jumps around
AI search readiness10%Whether AI crawlers are allowed in, and whether pages are written in passages an AI answer can quote
Images5%File sizes, formats, alt text

Three things this rubric does not score, and an honest audit says so. Off-site authority: who links to you and mentions you. Our 28 August audit noted that it "is not a scored category in this rubric, so the 69/100 understates the problem." Conversion: whether a visitor who arrives does the thing the page is for. That is a UX audit, a different document. Accessibility beyond alt text: contrast, keyboard use, screen readers. Each can matter more than everything in the table.

What five audits of our own site found

We relaunched suur.io on 13 July 2026 and audited it before launch and four times after. The figures below are historical observations recorded in our internal audit documents. Our weighted score is a prioritization tool, not a Google metric, a certification or a comparable score across every auditor. Changes in scope and measurement conditions can change the score even when the site has not changed.

10 July, before launch, on the staging build: 69. The homepage's mobile lab test showed 33 seconds of total blocking time: the browser's main thread was busy for most of the load. There were 22 doubled page titles and 7 missing canonicals. The old site had earned 17 clicks from Google in 16 months, and most of its top search queries were for a Dutch recruiter with a similar name.

29 July, sixteen days live: 64. The report recorded mobile PageSpeed scores of 59 to 62, with the largest element appearing after 8.3 to 9.4 seconds, and three search clicks for the relaunched site. Those were not the domain's first clicks: the earlier report recorded 17 for its predecessor. The score fell five points: content from 72 to 58 and AI readiness from 85 to 55. This audit assessed nine service pages as thin, noting 376 to 553 words each on a shared template. Length alone did not establish their usefulness.

7 August: 62. The report described the site as technically sound but commercially invisible. Its sampled checks found clean responses, canonicals, titles and schema across 286 URLs, while its keyword tool reported eight ranking keywords. It also recorded eight comparison pages with one internal link each, 92 single-post tag pages, 229 image files over 500 KB, and one ChatGPT response naming six competitors but not us. These were investigation leads, not proof of why rankings were low. The tool coverage and single AI query cannot establish complete search or AI visibility.

28 August: 69. This report measured a share-card issue on 138 of 275 pages, broader coverage than the earlier 29-page investigation. Its account of whether the previous audit had a complete score conflicts with the retained 7 August final summary. We therefore do not treat the score difference as a precise measure of improvement. It also found that of 30 searches near page one, 11 used site:suur.io: a poor basis for estimating customer demand.

19 September: 71. The report requested August 12–30 and August 31–September 18, nominally two 19-day windows, but recorded September 16 as the last available data day. These were not two complete, equally mature windows. It reported 1,868 versus 3,415 page impressions and 1 versus 12 clicks across the site. The 33 new URLs were a mixed cohort of service pages, articles and comparisons; that cohort had zero reported clicks, and 20 of the 33 checked URLs were indexed. These observations do not isolate publishing cadence as the cause or establish a commercial result.

What the score tells you, and what it does not

Three lessons from watching one site's score for ten weeks.

A score can change because the auditor looked harder. August's share-card finding covered more pages than the earlier investigation. If an agency shows you a before and after score, ask whether the same pages were measured the same way both times.

A clean score can sit on top of zero traffic. The 7 August site passed every hygiene check and was close to invisible. Hygiene is what stops a site from losing; it does not make a site win. What does is in the parts the score weights lightly or not at all: pages people are searching for, and other sites linking to them.

The results show up somewhere else. Read the findings and the Search Console section separately. An impression increase is an early visibility signal; clicks, qualified enquiries and sales answer different questions. The audit score does not substitute for any of them.

What a website audit costs

Initial evidence collection can use free tools. Our 28 September comparison recorded a 500-URL free crawl limit for Screaming Frog and a free verified-site allowance from Ahrefs; check their current limits before planning coverage. Google's PageSpeed Insights and Search Console also provide useful evidence. Collection time depends on site size, access and setup, while interpretation and manual testing take additional work. An afternoon is not a reliable estimate for completing a technical audit.

What you pay for is the reading, and the parts free tools do not do: the ranking of problems, the SERP and competitor checks, and a written plan. We priced the whole range, from free tools to agency retainers, in what website audit services cost, with every figure checked on the vendors' pages on 7 September 2026.

How to read an audit you are handed

Ask for the dated underlying evidence. For Core Web Vitals, distinguish lab runs from field data and identify the page and device conditions. For accessibility, W3C recommends combining tools with human evaluation. Neither a crawl score nor a clean automated check establishes that every user can complete the site's tasks.

  1. Find the top five fixes. If the audit cannot name them in order, it is a crawl report with a cover page.
  2. Check that each finding has evidence. "229 files over 500 KB" is an inventory finding, but still needs context about their use and measured loading cost. A reproducible broken keyboard interaction can be important without a score or percentage.
  3. Ask what was measured. How many URLs, which tools, which dates, lab or field data. A score without scope is not comparable to the next one.
  4. Look for what it did not score. Links from other sites, conversion and accessibility are the usual gaps.
  5. Look for Search Console. An audit that does not look at what Google actually shows your site is auditing a theory of the site.

Where this comes from

Our SEO and AI search audit is from $3,000, delivered in three business days, and the audit we ran on our own site is published in full as the sample audit, so you can read the deliverable before paying for one. If the question is the product rather than search, the Product UX Audit is $3,000 flat in three business days, and if it finds nothing worth fixing you pay half. For one page, the free page roast is the place to start, and hotels can inspect booking-link and metadata evidence with the hotel website grader. The performance finding that runs through all five audits became its own post: what WebGL costs a page, measured on this site.

A

Artyom Sklyarov

Founder & Designer at SUUR

More about Artyom →

Have a product idea?

Designed, built, and shipped in about 5 days.

Start a Project