We used Ai to create our website!
I recently encountered a local business that let me know they’d created their website entirely with Ai (I didn’t ask which agent designed it). I am not opposed to Ai use for a lot of tasks, but I am of the mind that all Ai websites are patently bad from an SEO perspective if a human hasn’t actually cleaned up the work. But this site had zero human-generated code, and even the content itself was also heavily created using Ai.
I’m not going to out the local business or even the industry they’re in, because hey, I suspect leadership there simply doesn’t know what’s going on and they aren’t paying me for diagnosing their issues. Ignorance is bliss, you know! But needless to say, their website is an absolute horror show that at the end of the day, is nothing but a somewhat broken, poorly indexed and badly optimized digital product offering an entirely unacceptable UX that isn’t doing very well in standard SERP, for any of the terms they could be dominating on.
First Clue: Poor Indexation
When I began my examination of the website, I wanted to know just what it looked like from Google’s perspective. So I did the usual on a Google search – site:theirdomain.com. Out of dozens of pages I saw in the flesh while visiting the website, only two (2) pages in total had been indexed by Google.
So from the get-go, this Ai designed and built website is hobbled from an SEO perspective. For key, high intent bottom of funnel searches – which should be bread and butter areas of attack as they’d obtain significant lead generation – the best the website did in Somerset and Pulaski county was page 9 of SERP. So “industry” or “business type” with or without the “near me” qualifier – nothing.
Almost every other local competitor with a website ranks above them. Even companies in the same line of business but well outside of the geographic and operational area rank better than them in local search.
A Redirect Chain of Horrors
Interestingly enough, rather than beginning the website content at the root domain https://theirdomain.com, the way Ai designed the experience to work is a redirect chain. First up, Ai designed the redirect chain to go from the root domain to the www subdomain via a 307 temporary redirect. And once you arrive at the fully qualified www.theirdomain.com, Ai decided it needed a 308 redirect to take you to https://www.theirdomain.com/home.
I’m sure you’re sitting there thinking – wait, what? A 307 followed by a 308 and then landing on what is the “home” page? Do they have some sort of API concern? As in is there a REST API that’s been redirected? Or is it a ton of legacy content that exists somewhere else, on a previous website?
No and No. There’s no logical or good reason for the chain.
A redirect chain like this – while possible, since hey, I see it happening – seems like an extremely questionable decision. I’m sure Googlebot is likely less than thrilled by this kind of technique. And from a strictly SEO perspective – the question this technique raises is if /home should be the canonical homepage URL rather than simply the root or subdomain.
End of day – I’d consider this approach to be a technical SEO flaw. Especially when “home” on the menu from any other page engages the redirect chain all over again.
Ai Designs (and Codes) Like Garbage
First check was to see if there were any robots.txt exclusions. And that was a negative. And so is there a no-index meta tag? Again, no.
Part of the issue with their site being indexed so poorly lies in how Ai decided to built it out. Quite simply, all of the HTML and content is hidden behind JavaScript. Bloated JS, at that. Unoptimized and loading just like molasses pours on a cold morning. There is zero content built in raw HTML – every single bit of the website is within JS and I suspect Googlebot goes in, sees an empty shell, and welp, decides there’s zero reason to index this page.
Site is built in Next,JS and is a Vercel site. And it certainly could be exported and delivered as static HTML, but I guess Ai didn’t let them know “hey, by the way, you could…”
For those wondering if a complete XML sitemap is presented, well, yes, indeed it is. Google still only indexed 13 percent of the website despite having a complete tree of all the content to plunder. And in terms of alternative search engines – the company does fare slightly better in local search on Bing, so let’s embrace those little victories.
E-E-A-T is Suspect
Because the content on page was Ai written, it often fails to agree with itself or offers up conflicts of information. IIn some places, the business guarantees they’ll call you back within two hours, maximum. In other places, the website insists they’ll call you back on the same day, every day. Conflicting information signals a lack of expertise and trustworthiness, period. And letting Ai write your content without correcting for blatant errors like this, not a grand idea.
One seriously problematic trust issue might be the displayed Google Reviews ranking. I’m assuming this was Ai’s doing – meaning they claim an aggregate of 4.8 on the website (with collated reviews that include misspellings of reviewers names – essentially typos I suspect? Does Ai make typos? Admittedly, I see it everyday in Ai where it changes things like that, but I digress.
Despite the displayed 4.8 aggregate rating, their actual Google Reviews aggregate rating sits at 4.7. I suspect there’s been a human element here – as the reviews appear to be curated rather than fed directly from Google.
But wait, because on pages other than the “reviews” page they go as far as to claim a perfect 5 star aggregate rating on Google – and then on the same page, further down, repeat the 4.8 aggregate rating. I get time – and they may have indeed had a 5.0 rating that then fell to 4.8 that then fell to 4.7 and they just haven’t bothered to update.
They also seem to cherry pick – only displaying perfect 5.0 ratings on the website as well, which also makes me lean towards human curation, but hey, I guess a prompt asking for only 5.0 reviews would work just fine, so…weird thing is on their review page, they offer displays of 5 star reviews in two different formats – a slider and a column layout on top of each other – but they seem to be the same 16 cherry picked reviews.
They also display an inaccurate number of total Google reviews, as their total is missing nearly 30 from what I currently see on the Google Business Profile which also suggests these were wholly human curations. Oh, and the whole total number of reviews – that seems to change page by page as well or in multiple places on a single page.
I don’t know much, but I do know enough to never, ever display information that would seriously destroy trust signals like this. E-E-A-T is a hard enough for companies without mucking it up by showing information that’s inaccurate and may add confusing signals to search engines.
DANGER WILL ROBINSON! DANGER!
– Robot, Lost In SPace
That said, in cases like this, where aggregate ratings are being inflated, I’d worry about the FTC knocking on my door, I’d worry about substantial hits to SEO across the board. And I’d worry about Google delisting my site and suspending my Google Business Profile.
And for those wondering – yes, there is indeed AggregateRating schema markup on page, meaning they’re essentially sending this error back to Google for inclusion in the SERP. That Ai tool does sure does know its schema, that’s for sure. No shortage of JSON, except for when it’s left out specific FAQs – which may indicate expansion of FAQs without asking Ai to create new JSON.
Finally, while I’ll delve more into the accuracy of information on their local listings, right now the website features an address that is completely wrong and miles away from where they’re actually located. If Google knows your website information is wrong, mainly because they have their own record of it in the Google Business Profile, well, maybe they can’t trust anything on this page at all?
Meta Title and Description, Headings and Thin Content
There’s something to be said about letting Ai write your content, since you know, you don’t have to do it, right? Only issue there is from what I’me seeing, the page titles lack target keywords they should be mentioning. The page meta descriptions are all too long and also don’t target keywords. Finally, both their on page and their location content is way too thin – meaning they have significant keyword gaps with the competition.
Ai also designed their RFI form as a two pager – meaning you have a natural, built in blocker to keep folks from successfilly completing a request for information. You enter a smattering of information and then hit a “NEXT” button. And boom, all of that information disappears and they ask you to submit more. I’m guessing from a conversion rate optimization standpoint, this means some leads fall off before they actually complete.
Accessibility? Pfft…No One Needs That!
A quick examination of the company homepage alone finds so many WCAG 2.1 errors I really don’t want to tell you a total. But they include: four (4) missing form labels and sixty-four (64) empty buttons, meaning no accessible names. Additionally, more than thirty (30) elements fail for sufficient contrast. The page is flagged with 79 other total alerts, including redundant alternative text, orphaned form labels, broken sane-page links, text that’s too small to read, HTML 5 video that lacks any text only format for those who are utilizing alternative means of access and zero transcripts for audio output. Finally, more than fifty (50) touch targets lack sufficient size and spacing.
Despite the numerous accessibility errors the Ai product has, I will say it did get keyboard only navigation perfectly right. So that’s certainly something, right? I shudder to think what might happen if an enterprising attorney that specializes in web accessibility gets wind of the website and finds a willing disabled participant that has been impacted.
Poor Lighthouse Scores
Since the site was completely coded by Ai, I thought we might see some promise, since, you know, a computer did the coding. So how does it do on lighthouse scores? Let’s just say those massive JS payloads and a huge DOM don’t bode well.
Take a look at mobile performance, for example:
And just for the record, there are a lack of items marketers like myself normally employ that impact the performance here, nowhere near what I’ve . This is almost all just the result of JS rendering everything and mucking it up (and hey, I was a Vue guy for a whole lotta years so I’ve made sites with heavy JS).
Note: even with a "perfect" 100 on SEO in lighthouse, well, this website's SEO performance is awful across the board. Moz identifies they only rank for a total of seven (7) keywords, all of which are three or four words long, indicating mid/long tail keywords and rank for absolutely no short-tail or head keywords at all, while the competition locally does almost across the board. So searches for /business type/ see them absent from SERP altogether.
While the Ai Website is Bad…I See Other Shortcuts Taking Place and Absolutely Dreadful Local SEO – Which are NOT an Ai Problem
There’s so much more to your local SEO performance than just your website. And the work on the other pieces of that puzzle also shows signs of taking shortcuts or less than optimal approach. First up – local listings. While they have a Google Business Profile and participate in Google Local Service ads, there’s evidence that they’re not on top of the other listings out there, be it from lack of claiming those listings, or simply not executing keeping listings updated to plan.
For example, their Birdeye listing features a different address and phone number than their actual business address and current phone number. Comparing listings and websites, I’ve seen a total of FOUR (4) different addresses being displayed. I get places of business move – so hey, we should reflect that, across the board. And the website itself – completely different address and NOT their physical address now.
And there’s almost zero agreement on times of operation across listings, with Apple and Yelp both indicating 24/7/365 hours of operation while the GBP pulls directly from their website and indicates they’re only open five days a week during normal business hours.
Having significant inconsistencies across the Having significant inconsistencies across the board in your business’s listings, such as name, address, and phone number (NAP), will confuse search engines like Google. This also splits authority signals – meaning your brand is represented as many weak entities rather than a stable, strong and singular business.
Conclusion: Shortcuts like Ai are Bad, Man
I’m only scratching the surface regarding the issues on the AI website, and I’ve noticed a similar lack of effort in local listings. However, theThe scattered inaccuracies and inconsistencies in the data, coupled with the complete disregard for E-E-A-T and the poor management of listings, make this business appear extremely low-quality. Which may be entirely unfair, but hey, life is often.
As a As a result, you’ll never achieve the coveted #1 ranking in the Local Pack (unless it’s an advertisement). Even worse, your content on your own property is likely to rank poorly, if at all, in search engine results pages (SERPs).
Deploying an Ai-designed website doesn’t guarantee failure or success in local SEO. But taking shortcuts across the board, including utilizing Ai to completely build a website and much of the content without ANY human work, with the added bonus of failing to do basic local SEO work, will guarantee a poor SEO result, impact your website’s ability to generate leads and sell your product/service.