Bing & Bard: the impact of AI search on website design

AI search is coming – and is set to be the biggest shake-up to search that we’ve seen in the last 20 years.

Right now, it feels like the only certainty around AI-powered search is how fast it’s changing and evolving. So fast that it’s hard to write about. By the time I’ve reached the end of this blog, it will have changed again.
But here’s my take on where we are now and the impact this is going to have on all of us as consumers, and on everyone whose job revolves around website design, SEO, advertising and content. People like me.
In short, the impact is going to be huge. Web search hasn’t changed this dramatically since the early 2000s when Google pretty much became search, and decided that, instead of categorising websites, it would rank them on how good they are.
And this new shift is happening because of the likes of ChatGPT. These large language models (LLMs) are deep learning AI algorithms that can recognise, summarise, translate, predict and then generate content, based on what it finds in huge datasets. The vision of Google (Bard) and Microsoft (Bing) is basically to stick one of these in between you and the search engine.
So instead of speaking to the search engine, we talk to an AI that uses the search engine to fetch what we need.
It’s a very human trait to get used to stuff and then to stop noticing it. Page ranking and our current version of search has been around so long, that we’ve forgotten the influence it’s had on our design and our behaviour. It’s quite rare that such a big technological change means we have to completely reassess how we do our work.
This is a radical moment. But how radical?

Users’ search expectations are likely to change

The first thing you’ll notice with both Bing and Bard is that you are prompted to ask a question in natural language, rather than the usual ‘search or type URL’, so straightaway you get the idea that you can do a lot more than search.
You can type in a multi-faceted question, eg, ‘What are 10 fun options for things to do, with my family, with kids, as tourists, in London?’ and it generates an actual, conversational answer, not a standard search results list. It delivers a ‘Here are 10 things you can do… etc, etc’. It answers all parts of my question and narrows it down to one full answer. Then it has a few links at the bottom.
So, the first really radical change is that the LLM is creating a search query based on what we ask. You ask a natural language question, and the LLM and Bing / Bard go off and construct their own search query based on that question.
This means we’ll have to change the optimisation we do on our side. It’s not for user queries directly anymore, but for what the LLM turns those into. We’re now optimising for AI search and what it thinks is a good answer.

There are very few links to websites

This change is huge, because the basic principle of the traditional search engine is that it takes you to find someone else’s website. It’s not doing that anymore. It’s giving you what it thinks is the answer based on other people’s content, and then gives you just 3 links below it. Bard gives a little more – an image with its ’10 things to do’, with a link from each image to the website – so it’s a bit more like a traditional list of 10 links.
But many users will get what they need without having to visit any website at all. Normally, Google or Bing helps us find a website to get the answer, but now they give us the full answer. Websites will consequently get a lot less organic search traffic, which will affect business models based on traffic. Media strategy will be affected too, because now you’re manipulating your traffic in a very different context. Another fallout from less traffic, is less data. And we all love data.
NB: Google’s Bard gives you those three pieces of content that it built that description from, but now it thinks of other related questions and issues, and starts to answer them. In short, it’s already predicting your follow-up questions.

SEO will need to work much harder

When you have just three links, SEO becomes more much more of a winner-takes-all situation. If you’re number 4, you don’t even show up, so it’s a very hard cut-off, and a rather brutal fist fight to get to the top of the list. Much more so than before.

Where users land on websites will evolve

Bing and Bard are going out, finding the specific content you need, and then assembling it into an answer that it then rewrites itself. And that answer is often on the destination or end page, not on an intermediate page like a category page, a landing page, or a list page. All of those intermediate navigation pages that we labour over, AI search just bypasses to delivers the answer. You can even perform tasks on the answer, still within search.
So, for example, I asked for 10 things to do with my family in London. Then I asked it to list the distance from the centre of London for each, which it did, all without leaving the search. E-commerce is the same. I can sort products by size, price, brand, all without a comparison tool.

What about accuracy?

Can answers be made accurate enough for Microsoft and Google to really stand behind them? At the moment, for example, below the prompt Google says ‘Bard may display inaccurate or offensive information that doesn’t represent Google’s views’. That’s a problem, because the UX is presenting it as Google’s single answer. My prediction is that they are going to back off a bit and may emphasise the source material more, therefore removing their liability.

But there is a plus side to all of this…

When a user finally ends up on a website, in theory they should be very happy to be there. So, while as a business you’re going to get less traffic, it should be of much higher quality. The final bounce rate of that last click should be really low, because they’ve already gone through a whole process. They’ve checked, they’ve validated and, when they show up to buy / book / sign up, they know it’s the right one.
And if Bing or Bard thinks your answer can be the answer, well that’s like being number one in the SEO ranking. But then again, exactly what that means is also still unclear. Everything is unfolding so rapidly and I fully expect to be revisiting all this on a weekly basis. What happens next is, frankly, anyone’s guess!