How Can I Optimise My Website For AI?
This question is on everyone’s lips at the moment. Well, everyone with a website who wants to get found by people at least.
I’ve recently added AIO (AI Optimisation) — also called AEO (Answer Engine Optimisation) — to my suite of SEO tools that I monitor and report on monthly for my clients. And it’s been a game changer.
Thanks to a great guy who is happy to share his learning with others, I got an early tip about how to go about this and it has given me, and my clients, an important leg-up in the age of AI.
First of all, what even is AIO?
Let’s go back a step and remind ourselves of what SEO is first. Put simply, SEO is the process of improving a website to make it rank higher on Google (what works for Google works for them all BTW). There are several elements to this:
Good content that is relevant for people who might be interested in your business.
Machine clues: these are page titles, descriptions and alt text on images that are not always seen by humans but help machines understand content more quickly (they also show up in search engines).
Functional websites: no broken links, good structure, sensible navigation and so on.
Backlinks: links from other reputable sites to yours.
Traffic: humans that visit your site and stick around for a while.
As for AIO, it works completely differently, darn it.
But you don’t have to learn a completely new skill or got back to school to master it. Most people use AI daily to answer questions on anything from “what hairstyle would suit me?” to “who’s the best carpenter around here?”
And if you’re a carpenter, you’re going to want to show up in the answer if at all possible. This requires something pretty easy for most humans to do: think like a human. Ask the sorts of questions humans ask. These are the steps:
1. Determine what your (or your client’s) business does or wants to be known for.
For example: a carpenter who specialises in bespoke kitchens.
2. If relevant, check whether you/they command an expert position on their subject.
Arguably, most carpenters will not have the time or the inclination to write blog posts or contribute to magazines about bespoke kitchens. But, let me tell you, a web designer like me definitely does. So cut your cloth accordingly on this one.
If this is relevant, you can ask AI: “Is Joe The Carpenter considered an expert in their field?”
3. Ask what the AI typically looks for when recommending [your type of business].
Your prompt: “When recommending carpenters who specialise in bespoke kitchens, what would you typically look for?”
At this stage, you don’t need to know if Joe will be recommended, just what the AI would be looking for.
Take the answer that’s given and check it out yourself. Look at their website, social media and anywhere else they appear online and check to see if the elements the AI is seeking are present. If they are, score a point, if not, no points.
4. Ask what AI already knows about [your business name].
Your prompt: “Tell me what you know about Joe the Carpenter.”
Read the answer and judge how close it is to what you (or your client) want to be known for. If it’s accurate, score a point for each element, deduct a point for every wrong answer. You might need to offer up the web address to help the AI pinpoint the right business (it is just a machine after all…)
5. Find out whether AI can discover brand mentions of [your business name].
Your prompt: “Search everywhere online except joethecarpenter.com and tell me what you find. I’m interested to find out what others think of this business.”
See how many mentions come up and score a point for each. You may need to give additional prompts such as where to look and be aware that some sites prevent AI crawlers (LinkedIn for example).
6. Ask what the sentiment is for [your business name].
Your prompt: “Having seen these third party mentions of this business, what would you say the overall sentiment is?”
Score a point for all the positives and deduct points for the negatives. Use your judgement about this.
7 Finally, would AI recommend [your business name]?
Your prompt: “Given everything you now know about Joe the Carpenter, would you recommend this business to build me a bespoke kitchen?”
Score a point for yes and anything resembling yes, deduct for no.
Tally up your scores for each element and turn that into a percentage. For example: If the AI finds three positive reviews and three negative that’s a score of 50%. But you can use your humanity. So if the AI says: “There are three really good positive reviews and three that are tending a little negative.” you might want to weight your scores. Similarly, if the negatives have all been responded to, you’ll notice the AI will give credit for that. So a straight 50/50 might turn out to be 60/40.
What’s the point of all this?
The point is to get your scores as close to 100% as you can. And you will end up with very clear actionable ways to improve because the AI will tell you.
For example, when asked, what do you know about Joe the Carpenter, the AI might say: “I can see they are an established business and they have a few good reviews but I can’t see any evidence of insurance or membership of trade bodies. These are things I would definitely want to see before feeling able to recommend them.”
Most carpenters would have both. Simply pop them into the website and recheck (not immediately, give it a week or so).
And I know this works, because I’ve done it for many clients.
There’s a handy guide to help you, created by my friends at SEMrush. I use SEMrush to manage all my client SEO and AIO work. Their dashboard includes a high-level overview of AI visibility too which is a quick reference I check every month.
