How to Run an AI Website Audit in 3 Passes
Paste your URL into AI and it tells you your messaging is wrong. Give it your real analytics and the diagnosis can change completely. Here are the three passes, and why the last one is the only one worth acting on.
The short answer
An AI website audit works in three passes, and the answer changes at each one. Paste your homepage URL into a brand new chat with no context, and it tells you what a stranger sees in ten seconds. Move it into a project and ask for a brand audit, and it finds the story your site is burying. Then give it your real analytics, and the diagnosis can change completely. On the store I audited, pass one said the messaging was wrong. Pass three said the business had nearly stopped functioning. Same site, same tool, and the only difference was the data.
At a glance
- Time
- ≈ 30 minutes
- Cost
- Claude Pro
- You need
- Your analytics
- Where
- A brand new chat
I asked AI to do a full audit on a real e-commerce store.
It came back with a list of messaging problems. Wrong headline. Buried value proposition. Weak product descriptions. All annoying, all fixable.
Then I did one more step, and the story completely changed.
I can't show you the store or the niche, because when you put a live website on screen some people go and do unpleasant things to it. It doesn't matter. The method is the same whatever you're running, and so is the strategy.
You have an online presence. You want to know where you're at, what needs fixing, and in what order.
That last part is where most of this goes wrong. There's always more you could fix than you have time for, so the priority order is the whole prize.
Watch the video above to see all three passes run on the real site.
Key takeaways
- 01
Start in a brand new chat with nothing but the URL. No personalisation, no context, no explaining your business. You want the read a stranger gets, and any context you add is context a first-time visitor never had.
- 02
The first pass is observation, not diagnosis. It's smart, it's often uncomfortable, and it has no data behind it. Treat it as a list of things worth checking rather than a list of things to go and fix.
- 03
Real analytics change the answer itself, not the detail around it. The messaging problem turned out to be sitting on top of a traffic problem, a tracking problem, and months where the store made nothing.
- 04
Push back when it recommends something. I challenged the recommendation twice, and both times it either defended itself with a reason or admitted it was working without data. Both outcomes are useful.
- 05
Fix where the money is collected before you drive more traffic to it. As I put it in the video: "fix what happens when someone arrives at the website, then drive more people to arrive."
Here's what pass one gave me from nothing but a homepage URL.
It worked out the niche, that the store runs on WooCommerce, what it sells and who it's for. Then it started listing what it saw. Last site update April 2023. Hero image from 2015. Logo from 2014. Navigation images untouched since the site launched.
No brand differentiator. No reason to stay. Prices covering a decent range with nothing anywhere explaining why you'd buy here rather than somewhere else. No trust signals, no social proof, no why us.
That's brutal, and it knew nothing about this business except the address.
But every bit of it was still observation. Nothing there was measured. That distinction is the thing I want you to take away, because it's what separates a useful audit from an expensive month spent fixing the wrong thing.
Prefer to watch
12:11The three passes, in order
Paste the URL into a brand new chat
Open a completely new chat. No personalisation, no context, nothing about your business saved anywhere in it.
Paste in your homepage URL. That's the whole prompt.
If your AI is already set up to know you, that's normally an advantage and here it's a problem. You want the cold read, the one a stranger gets.
Read it without defending anything
Everything it says at this stage is an observation from your public pages. Let it be.
Write the list down. Don't argue with it yet, and don't start fixing anything either, because you don't yet know which of these actually costs you money.
Move it into a project and ask for a brand audit
Same lack of context, one structural change: put the site into a project and ask for a proper brand audit.
The output gets much sharper. It'll look at your banner, your logo, your photography, your product pages, your about page, your socials, and it'll start naming patterns rather than faults.
Ask it for competitor research from here too. That pulls conversion and reputation into the picture.
Ask it for the losing moment
This is the question that earns its keep. Ask for a report covering positioning, audience fit, trust, and the losing moment. Where does the site lose the customer?
Then ask what the real problem is underneath all of it.
Push back on whatever it recommends
Mine told me to focus on the website. I argued, and asked whether we should just get some Instagram Reels going instead.
It said no, and gave me its reason: there's traffic already arriving at the site, and the socials aren't sending anyone anywhere.
Then it handed me headline copy, and I pushed again. You've got no real data. This is you reading a website.
It agreed. The audit was observation and competitive analysis, with no numbers in it at all.
Give it your real numbers
Add project instructions explaining what the business actually is. Then add the data. Google Analytics, your socials, who's doing what.
I pulled the analytics myself and dropped them into the project folder. You can get AI to pull them for you, but grabbing them by hand takes minutes and removes a step that can go wrong.
Take the priority order and start at the top
Now you get an order based on what the numbers say will move the needle, rather than what looked worst on the page.
There is always more to fix than you have time for. This is the part that decides whether the next month of your effort counts.
What each pass actually found
The banner doing nothing
The most valuable space on the homepage, and it was floating brand logos. No offer. No reason to stay.
The logo works, but says nothing. The whole site read like a catalogue rather than a company, right down to photography that was flat product shots.
The real story, buried
This is a family owned business. That's a genuine advantage and it was sitting in the second paragraph of the about page.
Product pages were catalogue style, SKU and name, thin descriptions, barely any features. A few broken pages. Almost no reviews.
The selling point hidden in the footer
Here's the one that stung. It's a Canadian store, prices are in Canadian dollars, and duty and customs are already covered for a Canadian buyer.
That's a real, concrete reason to choose this store over a competitor.
It appears nowhere on the site except fine print in the footer.
What changed when the analytics went in
Same audit, completely different story.
The traffic was mostly smoke, with a lot of it bots. It told me what the store actually made, and there were months it made nothing.
Google Analytics was running but had never been set up to track the e-commerce side, so there was no view of what people did on the site or where they dropped off.
One brand page was carrying almost all the real traffic. The homepage was invisible in search, the keywords weren't hitting Google at all, and one of the few things still pulling people in was a service the business doesn't even offer any more. It sent them nowhere.
Pass one said the messaging was wrong. Pass three said the store had nearly stopped functioning as a business.
The priority order the data produced
First, fix the e-commerce tracking and Google Analytics, so from that day forward there's a real picture of what people do on the site.
Then the service page redirects, so the traffic still arriving lands somewhere that sells something.
Then the product pages. Then reviews and socials, last.
Note where reviews ended up. In pass one they were the top priority. Nothing about the site changed. The data just told a different story about where the money was going.
Sources
Questions people actually ask
What do I need to run an AI website audit?+
Your homepage URL and a chat you haven't used for anything else. That's pass one. For the later passes you'll want a project and your analytics export, which is where the audit stops guessing.
Why start with a brand new chat?+
Because you want the read a stranger gets. If your AI already knows your business from memory or custom instructions, it fills in gaps your visitors can't fill in, and the audit comes back kinder and less useful.
Can AI audit my social media as well?+
Yes, and mine did it in pass two without being asked separately. It found YouTube effectively dead and Instagram and Facebook present but doing nothing, which matched the pattern on the site.
Why does the audit change so much when you add analytics?+
Because everything before that is a reading of your public pages. It's a smart reading, but nothing in it is measured. Numbers tell you which of those observations is actually costing you, and that reorders the whole list.
Should I fix my website or my socials first?+
The site, in almost every case, and here's the reasoning it gave me. Fixing your socials before fixing your site is like running ads to a broken store. Fix what happens when someone arrives, then work on getting more people to arrive.
What if my analytics were never set up properly?+
Then that's your first job, and it was the top item on this audit too. Tracking that doesn't track your sales means every future decision is a guess, so it comes before any cosmetic fix.
How far back should the data go?+
Far enough to see the pattern. This store launched around 2014, put its digital presence together in year one, and coasted afterwards. No meaningful update since about 2020, which is 6 years of drift by the time I looked at it.
What do I do with the priority order once I have it?+
Start at the top and don't skip. The order is the output. There's always more you could fix than you have time for, so the value is in knowing which single thing moves the needle first.
About the author
Hey, I'm Traci
Digital Marketing Strategist
With over 25 years in this industry I help small business owners and creators turn their online presence into profit.
Keep going
The setup, the templates, and what I build next
Not more reading. The Claude setup guide you can follow tonight, and the templates I actually use. The page opens the moment you sign up.
Get the guideFree · you get it straight away