One URL, No Login, Nothing Found
Why I call it the last mile
In shipping, the last mile is the final stretch between the warehouse and somebody's front door. It is the part that costs the most and gets the least attention, because it is unglamorous and it is nobody's favourite job.
I borrowed the phrase because it describes what I keep finding on small websites. The hard part is usually done. Somebody built the site. It works. What is left is the short, dull distance between a site that works and a site that works for everyone who tries to use it — and that stretch tends to sit untouched, not because it is difficult, but because nobody enjoys it.
This post is about the least intimidating way I know to see that stretch on your own site.
The number at the top is not the point
The tool is Google's PageSpeed Insights. You type in one web address. There is no account, no card, no cookie banner, no extension to install. It gives you back two things at once: how quickly the main content shows up, and a list of accessibility problems it can detect.
It also gives you a score out of one hundred, in a coloured circle, at the top of the page.
Skip it.
I mean that. The score is the first thing you see and it is the least useful thing on the page. A high number means the automated check found nothing it knows how to look for. It does not mean a person using a screen reader can get through your booking form. Those are different questions, and only one of them can be answered by a machine.
Scroll past the circle. Underneath it there is a list, and the list is the whole value. It names things — this image has no description, this text is too faint against its background, this button has no label. Each item is a specific thing on a specific page. That is something you can act on, or hand to someone, or ignore on purpose. A number is none of those.
I would rather you never screenshot the score at all. The moment a number like that gets shared, it starts to look like a verdict, and it is not one.
What came back, and what did not
I ran it on my own site, which is the only site I have any business showing you results from.
It found nothing. Zero failed checks on the accessibility side.
I want to sit on that for a second, because the wrong lesson is right there waiting to be picked up. The wrong lesson is that my site is accessible. What actually happened is narrower and much less flattering: an automated tool looked for the things it knows how to look for, and did not find them.
Google says as much on the same page, in its own words. Automatic detection can only detect a subset of issues, and does not guarantee the accessibility of a site. That sentence is printed directly under the result. It is easy to scroll past, and it is the most important thing there.
Then it did something I think is genuinely useful. It handed me a list of ten things it cannot check, by name. Whether every interactive control can be reached with a keyboard. Whether the tab order through the page is logical. Whether the visual order matches the underlying order. Whether focus ever gets trapped somewhere a person cannot escape. Whether content that is hidden off screen is properly hidden from a screen reader too.
Read that list again and notice what it is. Every one of those needs a person. Not a better tool, not a paid tier, not an upgrade. A person, with a keyboard, going through the page.
So the honest summary of my result is: the machine found nothing, and it told me plainly that there were ten places it did not look. That is a floor, not a ceiling. It is still a concrete place to start, and starting is the part most sites never do.
One more thing worth saying out loud: the figures that get quoted about accessibility describe the web at large, or the top million home pages. Nobody has measured small business sites specifically. If a site is small and quiet, it is arguably excluded from those datasets rather than represented by them. So I will not tell you what percentage of shops like yours have a problem. I do not know, and neither does anyone selling you a number.
Where this fits if you are new to AI tools
Most of what I do is help people start using AI tools like Claude, ChatGPT and Google's AI, without the fear or the hype.
The thing I keep coming back to is that this kind of work makes a relatively safe first practical use case. Relatively. The stakes are low, the results are checkable, the mistakes are visible and reversible, and the work is boring in a way that makes it hard to break badly. You can ask an AI tool to explain what an item on that list means, and then go and look at the actual page to see whether the explanation holds up. That loop — ask, check against reality, correct — is the whole skill, and this is a gentle place to practise it.
Compare that to asking an AI tool to write something you will put your name on, or to make a decision with money attached, and the difference in risk is obvious enough.
The sources of truth here are WebAIM and the W3C, not me. If you want to read further, go to them directly. I would rather point you at the thing itself than at my own site.
And if you run the check and the list makes no sense, that is the normal outcome, not a failure on your part. It is written for developers. Text me at 503-664-0546 and I will read it with you, no charge and no pitch. If texting is not your thing, eog@ernestofgaia.xyz works, and we can set up a call from there.
Reading it with you is free. If it turns out the list is long and you would rather not spend your evenings on it, that is a separate conversation and we can have it then.
#ernestGoesToAI