What an audit is actually for
The point is to decide what to do next. That is the only test that matters. If you finish reading and still do not know what to do on Monday, it failed regardless of how thorough it looks.
Watch out for length being sold as rigor. A 200-item report usually means somebody exported a tool and did not do the work of separating what matters from what does not. Ten findings you will actually act on beat two hundred you will not.
Can Google find and read the site
First question, always. Not how the site scores — whether your important pages are being found at all.
Check which pages are actually indexed against the list of pages you care about. Look for the ones that are missing. Dealer sites are very good at burying the service page, the finance page and model pages behind scripts, menus and thin navigation.
Also check the reverse: thousands of near-identical inventory and filter pages soaking up attention that should be going somewhere useful.
Inventory and templates
Your platform generates pages from templates, so a flaw in one template is a flaw on thousands of pages at once. That cuts both ways — it is also why one fix can move a lot.
Look at what happens when a vehicle sells. Does the page vanish, redirect somewhere sensible, or sit there as a dead end? Dead ends at scale are a real problem and most stores never look.
Check that filtered and sorted versions of inventory are not all competing as separate pages, and that titles and descriptions are not identical across hundreds of listings.
Local and department listings
This is where dealer sites lose most often, and it is the cheapest thing to fix.
Your name, address, phone, hours and departments should be identical everywhere they appear: your website, Google Business Profile, manufacturer locator, directories. Down to the abbreviations.
Check each department has its own presence where it should — sales, service, parts, collision. Check the hours are right, including holiday hours nobody updated. Check somebody is answering reviews and questions.
Speed, and whether people can actually do anything
Speed matters, but not as an abstract score. What matters is whether the page is usable quickly on a phone on a normal connection, because that is how most people arrive.
Then check the things people came to do. Can they book service in under a minute? Does the finance form work on a phone? Is the phone number tappable? Do the buttons work?
A dealership site that scores well and cannot take a service booking on a phone is failing at the only thing that counts.
Put it in order of what to fix first
Findings without priority are noise. Every item should carry two things: roughly what it is worth, and what it takes to do.
That produces the only useful output — a short ordered list. Fix this now because it is cheap and blocking. Fix this next because it is worth the most. Ignore these because they cost more than they return.
Be honest about the third category. Some findings are technically true and not worth anyone time, and saying so is part of the job.
Verify the fixes actually landed
This is the step everyone skips. Work gets marked complete, nobody re-checks, and six months later the same finding shows up in the next audit.
After changes go in, re-run the specific checks that failed. Confirm the buried pages are now indexed. Confirm the listings match. Confirm the booking flow works on a phone.
An audit nobody verifies is a document, not an improvement. If your vendor never re-checks their own work, that tells you what the audit was for.