Generating a website is the easy part now. Any tool can produce something that looks like a website in under a minute, and a good many of them do.
Whether it works is a completely separate question, and it is the one that decides whether a buyer rings you. A dealership site has a job list that a restaurant site does not: it has to carry live stock, take an enquiry that actually arrives, handle a finance application without losing the applicant, and stay readable on a phone in daylight. Generated quickly and checked not at all, a site can fail every one of those while looking perfectly smart in a screenshot.
So there is a checking stage between generation and you. Here is what it is for.
Checked against what the job actually requires
Every generated site is checked against what a working dealership website has to do. Where something is wrong, it is corrected and checked again, before the site is ever shown to you.
That means the structural things, which are also the boring things nobody photographs. Does every vehicle have a page and does that page carry its photos, price, mileage and specification. Is there a way to enquire on every car. Does the finance application run in steps a person can finish. Can a buyer shortlist vehicles and compare them. Do your contact details exist and are they where a buyer looks for them.
These sound too basic to need checking. They are exactly the things that go missing when a design is generated from a description, because a description is about how a business feels and a checklist is about what a page contains. The two have to be reconciled by something.
Told in plain English
When a check cannot be satisfied, you get told what the problem is in a sentence, not handed a list of faults to interpret.
This is a deliberate choice and it is worth explaining. The usual output of an automated site audit is a report with forty items on it, most of them written for developers, ranked by a severity scale that means nothing to a dealer. What a dealer does with that report is nothing, because the first item is incomprehensible and there are thirty-nine more behind it.
A problem described in plain language is a problem you can make a decision about. That is the whole test.
Your colour, still readable
The check that comes up most often is contrast, and it is worth understanding because the fix is not the one people expect.
Brand colours are chosen on signage, on a printed banner, on a van. They were never chosen for small text on a screen. A mid-tone colour that reads beautifully on a forecourt sign can be genuinely hard to read as a button label or a heading, particularly on a phone held at arm's length outdoors, and particularly for the substantial number of buyers with any degree of colour vision deficiency or simply middle-aged eyes.
The accepted standard is a minimum ratio between text and the surface behind it. Below it, some proportion of your buyers cannot comfortably read your call to action, which is an expensive place to have a legibility problem.
So if your brand colour would be hard to read against the theme you have chosen, the site says so, and offers a nearby shade that works. Not a different colour. A version of yours, adjusted enough to be legible, close enough to still be recognisably you. You can decline it. You should know before you decide.
Thorough, not infallible
That phrase is on the feature page deliberately, and we would rather say it here too than have you find out on your own.
Automated checking catches categories of problem. It catches missing pages, broken structure, unreadable colour combinations, things that are measurably wrong. It does not catch things that are only wrong if you know the business.
It cannot know that your opening hours changed in March. It cannot know that the headline describes the dealership you were three years ago. It cannot know that the photograph of the forecourt shows the old signage, or that you no longer offer the warranty the about page mentions, or that the marque you specialise in is not the one that leads the page.
None of that is a failure of checking. It is the category of thing that only the owner can verify, and no amount of automation moves it.
What is still your job
Before you publish, five minutes on this list, on your phone rather than your desktop:
- Read the headline out loud. If you would not say it to a customer, change it. This is the single most common thing dealers leave as generated.
- Check the facts. Opening hours, address, phone number, the services you actually offer, the warranty you actually give.
- Submit an enquiry yourself, on a real vehicle, and confirm it reaches you and is attached to that car.
- Click into three vehicles at random and check the specification looks right.
- Look at it outdoors if you can. That is where your buyers are.
Then publish. And if something is wrong after you do, an earlier version can be restored in a click, which is the reason to be relaxed about pressing the button rather than the reason to skip the list.
Why the safety net changes how you work
The rollback is worth a paragraph on its own, because it quietly changes the relationship dealers have with their own site.
Most dealers stop editing their website within a month of getting it. Not because editing is hard, but because they changed something once, it looked wrong, and nobody could tell them how to put it back. After that the site becomes a thing you do not touch, and it slowly drifts away from the business it describes.
If getting back is one click, that fear goes, and you keep the site current. That is worth more over a year than any individual design decision.
The short version
Fast generation and careful checking are not the same feature, and a tool that only does the first one is handing you a liability with a nice hero image on it. What arrives should already have been checked, what could not be fixed should be explained to you in language you can act on, and the last look should still be yours.
See it on your own stock at the AI theme builder, or read the launch post for how the whole thing fits together.
