In this article
Search is often treated as a website job. Someone looks after the pages, someone reports the traffic. AI search makes it a growth question.
Someone choosing a supplier, a service or a charity to support can now ask an AI assistant and get back a few names. Google puts AI answers above some of its own results too. By the time anyone visits a website, part of the shortlist may already be made.
The website side of this, what makes a page easy to find, read and cite, is in what AI search means for your website. This guide covers the rest: brand, content, proof, measurement, and what to ask for.
How a shortlist gets made now
An assistant draws on a range of sources before it answers. Google says its AI answers can break one question into several related searches, so an answer may link to a wider range of pages than a classic search would. Your website is one source. Reviews, directories, articles, news and what other people say about you are others.
That matters for growth in two ways. First, a shortlist can form without you. Someone who reads an answer naming three of your competitors may never search for you, and a visit that never happens doesn’t show up in your analytics.
Second, the people who do get in touch arrive with a view of you already formed, from a summary you didn’t write. Some of the selling, good or bad, has happened before the first conversation.
None of this needs a new department. It does mean that how you are described, by you and by everyone else, is now part of how you are chosen.
Brand: the answer is only as clear as what is said about you
If your website says one thing, your LinkedIn page another and an old directory listing a third, an answer may blur them or pick the wrong one.
An assistant can only be as clear about you as you are about yourself.
The problem is older than AI. National Energy Services had several established brands, divisions and websites, strong individually and confusing as a whole. We brought them under one name, Sava, and one corporate site that explained the offer to each audience. That was done for people, not machines, but the fix is the same: one clear account of who you are.
So ask whether everyone who talks about you could say, in one sentence, who you are for and why you are the right choice. That is positioning, and what is brand positioning? covers how to find yours. An offer that is harder to explain than it should be is hard for an assistant to explain too.
Content: answer the questions the whole decision asks
Assistants answer questions, so they draw on content that answers them well. The leadership point is whose questions. Many decisions involve several people, each asking something different: the person who will use what you offer, the one who signs it off, the one who has to defend the choice later. Marketing to a buying committee shows how to map them.
In workshops with Wellcome Connecting Science, we grouped the website’s audiences into three, from researchers and funders to educators, the public and the media. We framed each of their tasks as a single job to be done, then drew a route for each group from the homepage to where they wanted to be, with as little effort as possible. That map is also the list a content plan needs: who is asking, and what they are trying to get done.
Google’s guidance asks for a point of view of your own, such as first-hand experience, and steers away from content a generative AI model could easily produce. An article anyone could have written gives an assistant nothing it needs. A few pieces built on what only you know are likely to do more than a steady stream of filler, which makes this a budget question as much as a writing one. The content that actually earns citations shows what that looks like.
Proof: give people something they can check
Proof used to work on whoever reached your case studies. Now some of it works before they arrive. Google says its AI answers can show what people say about products and services across blogs, videos and forums, and that chasing fake mentions isn’t as helpful as it might seem. So the job is the old one, done properly: work worth talking about, and clients happy to say so in public.
Proof also has to suit the reader. Anderson Group, a global outsourcing and payroll business, was selling to people with little time or interest in the subject. So we made short films, case studies, testimonials and a suite of animations to debunk the myths of outsourcing. Material like that answers a doubt before anyone raises it.
Figures need the most care, because a number repeated in an answer can travel without its context. Give each one a source, a date and a scope wherever you publish it. Social proof that works has more on where proof does its job.
Measurement: count what you can, and say what you can’t
This is where a leadership team most needs honesty from whoever reports to it. Some of AI search can be measured. Some can’t yet.
| What you want to know | Where it comes from | What it can’t tell you |
|---|---|---|
| How you are found in Google | Search Console, Google’s own figures for your site | Anything about other assistants |
| Whether you appear in Google’s AI answers | Search Console’s Generative AI performance report, which counts appearances in AI Overviews and AI Mode by page | What the answer said about you |
| How other assistants describe you | Your buyers’ questions put to them each quarter, the same ones each time, with the answers written down | A figure you can trend with confidence |
| Whether it turns into business | Enquiries, and asking each new enquirer who else they looked at and how they found you | Why you missed a shortlist you never heard about |
For the sites we look after, Search Console, Analytics and the searches we track come in every Monday morning, and the figures are worked out from the data, never typed in. Each quarter we look at how AI search results cite and summarise each site, write down what to change, and report it as we find it, not dressed up as a ranking. There is more in SEO reporting and AI SEO.
What a leadership team should ask for
- 01One description everyone uses
Who you are for, what you do and why you, in one sentence, the same on your website, LinkedIn, listings and bids.
- 02Proof with its working shown
Case studies and figures that carry a source, a date and a scope, and the client’s yes for every use.
- 03Reports that keep measured and estimated apart
Google’s own figures first, anything estimated marked as such, and gaps shown as gaps.
- 04A decision on AI crawlers, made on purpose
Whether assistants may read your site, decided by someone senior and written down so it isn’t changed by accident.
Be wary of anyone who guarantees you a place in AI answers. Nobody controls what an assistant says, and Google says no outside tool can see inside its ranking or AI systems.
If a second pair of eyes would help, our free marketing audit comes back with three things you could fix this month. No pitch, no tie-in. You can read more about how we use AI, or just get in touch.
Common questions
Will AI search cut our website traffic?
For some quick questions it can, because the answer arrives without a click. People making a considered choice still visit, often after reading a summary first. So judge search by enquiries and their quality, not by visits alone.
Should AI search change our marketing budget?
Probably not the total. It may change the mix, with more care over proof and over how you are described elsewhere. Google calls optimising for its AI answers “still SEO”, so it rarely needs a line of its own.
Who should own this?
Whoever owns how you are described. That is often marketing, with the leadership team agreeing the one-sentence description and the proof you will stand behind. A web team looks after pages. It can’t fix a brand that says three different things.
Does this apply to charities and the public sector?
Yes. Someone deciding which charity to support, or which service to use, can meet the same short list of names. The need to be clearly described, with proof people can check, is the same.
