A free website audit that builds itself, and still waits for an engineer
Our free website audit needs only an address. The checks, the summary email and four documents are built automatically. Before anything goes out, an engineer approves building it and then approves sending it. Here is how it works, and why both halves matter.
A free audit is only worth offering if it is fast, consistent and correct. Doing every one by hand is slow and varies with whoever does it. Fully automating it is fast, but then a machine decides what a stranger reads about their own business, with nobody answerable for it.
We built ours to get both. A machine does the work. An engineer owns what is sent.
If you lead the team: what to ask
- When our tools send something to a customer automatically, who checks it first, and at which step?
- Does every claim in an automated report come with evidence the reader can check?
- When the automation can’t finish, does the request reach a person, or does it disappear?
One input: the website address
You give us your email and your website. That is all the audit needs. It only uses what any visitor or search engine can see: no logins, no security scans, and nothing on your site changes. What a free audit can see from the outside explains why that is already a lot.
From the address, the tool makes about 50 requests to the site and runs Google Lighthouse (Google’s own quality test for web pages) on mobile and on desktop. Then it answers eight questions:
- Can search engines find and list your pages? (robots.txt, sitemap,
noindexand canonical tags, titles and descriptions) - Do old addresses,
http://andwww, lead cleanly to one site? - How fast is it on a phone and on a computer?
- Is your analytics recording?
- Can visitors and search engines reach the site at all?
- How do AI assistants and link previews see it?
- What does Lighthouse flag for accessibility and best practices?
- The basics: mobile layout, broken links, security headers, a site icon.
Evidence, not opinions
A finding is only raised on something the tool actually observed: a status code, a header, a tag, a rule in robots.txt, a measurement. Each finding carries that evidence, so you can check it yourself in a minute. The wording is fixed and written by us for each kind of finding, in English or Spanish, and the evidence is filled in. No AI writes the findings, so a report never claims a cause nobody saw or a figure about your business.
Findings are ranked by what is most likely to cost you: anything that hides pages from Google or blocks people comes first, then broken measurement, then speed, then the rest.
What you receive
- A short summary email with every finding, most important first: what it is, why it can cost you, how to check it, and the first step to fix it.
- Four documents: the website check with all the evidence, the Lighthouse results for mobile and for desktop, and the keywords on your pages, judged from the pages themselves.
The documents come as PDF attachments and as private links, so you can forward them to your team, your agency or your developer as they are.
The free tier runs only on free resources: Google’s free speed-test service, the mailbox we already had and the storage included in our hosting plan. That keeps it free to offer, with no meter running behind each request.
Two approvals before anything goes out
Automation could send the report the moment it is ready. We chose not to let it. Every request waits for two decisions from our team:
- Approve building. A person sees the request first. A real site gets built; a typo, a test or an obvious misuse doesn’t.
- Approve sending. The finished report waits for review. An engineer reads it, then sends it or rejects it.
Why the second look, if the checks are reliable? Because a correct finding can still be the wrong thing to send: a site that was mid-migration that day, or a page that was meant to be hidden. The tool can’t know that. A person can tell, and the person who presses send is answerable for it. If the tool can’t reach a site at all, the request doesn’t fail quietly: it goes to the team to handle by hand.
Every audit is kept as a record, so if you ask a question about your report later, we are looking at the same report you received. Requests are limited to a few a day per email address, which keeps the tool there for the people it is meant for.
This is how we use AI and automation in client work too, and it is the subject of AI speeds the work; people stay responsible: the machine does the repetition, people own the result.
Tested on ourselves first
Before offering it to anyone, we ran it on evdevs.com. It found real problems on our own site, and we fixed them: here is what it found and what changed. The full path, from request to approval to a report with its PDFs and private links, was tested end to end on the live site before launch.
What happens after you ask, step by step, is in what happens when you ask us for a free website audit. No call, no spam.
Or read what the free audit checks first.
Checklist: automating a report you send to customers
- Does every statement in it rest on something you observed, shown as evidence?
- Is the wording fixed and reviewed, or written fresh each time by a model nobody checks?
- Is there a person who approves before it is sent, and do they know they own it?
- When the automation fails, does a person get the request?
- Do you keep a copy of exactly what each customer received?
- Is there a limit, so one person or one bot can’t run up your costs?
Written from our engineers’ work on production systems. Want a second opinion on your project? Talk to an engineer.
See the work →