Pay attention to shifts in search intent

Mark says: “You should be monitoring search intent shift. Be aware of when what users want, and what they mean when they use a search term, changes over time – and adapt accordingly.”

How do you find out what users want?

“That’s the million-dollar question. Intent shift happens in several different ways. Some are ways that people are already aware of, and others are harder to recognise because they are happening in the background.

I like to think of a search term as a spotlight, and there is a universe of things that people want to find out. That search term spotlight slowly moves around this universe of information as the intents shift.

Take Halloween, as an example. If you carry out searches related to Halloween throughout the year, you get a lot of informational sites describing the origin of Halloween, the background, and what it’s about. When you start coming up to October, searches for Halloween spike and the results are a lot more commercial in nature. They start showing e-commerce sites that sell Halloween costumes, candies, etc.

That’s because Google has monitored (as we now know, after the DOJ trial) how users are interacting with the SERP. They do so in a predictable fashion, and the intent of what they actually want changes.

The other way that I monitor search intent shift is with People Also Ask data. People Also Ask data gives you an ‘intent proximity’. When someone searches for a phrase, a question, or even a single keyword, the People Also Ask questions give us insight into what Google thinks is going to be helpful as a follow-up question.

That’s so important because one of the signals that Google uses to monitor user satisfaction on the SERP is whether they go back and redefine their query. If someone searches and then goes back and redefines their query, Google gets a hint that they weren’t showing the right thing the first time around. People Also Ask data is a hint as to what Google thinks they might look for next.

As part of what we do at our agency, we monitor how People Also Ask questions change over time. We’re scraping them, storing them, and seeing what Google’s suggesting. If you look at the People Also Ask questions related to COVID, two years ago, they were quite panicked: ‘What do I do if I have COVID’, ‘What are the symptoms?’, ‘What is the law?’, etc. Now, two years later, people understand what COVID is, and the intent has shifted from trying to find out what it is to, ‘How do I manage my symptoms and get on with my life?’

That is a big signal across any industry you’re working in. If you see changes, and the spotlight of that search intent is moving, you need to update your content to match the current intent. With COVID, it’s shifting in one direction because our collective knowledge is changing. With Halloween, it’s oscillating because the intent changes based on the time of year.”

Are there other ways that search intent can change?

“One of the things that was absolutely fascinating in the DOJ trial, was that Google talked specifically about a system they called Instant Glue, which is a version of NavBoost that looks at and remembers how users are clicking and adjusting SERPs.

They gave a very intense example in the trial. If someone is doing a search for ‘nice pictures’, Google will give them nice pictures. However, if there was a terrorist attack in Nice and people searched for ‘Nice pictures’, Google would realise that the way people are interacting with the SERP has completely changed. They might be clicking on news websites or images of Nice, the city, rather than ‘nice pictures’. That’s how they know when to readjust the SERP.

While there aren’t a lot of things you can do to prepare for that, because you can’t tell when that shift is going to happen, it’s worth knowing that those mechanisms exist, and Google is thinking about that.

I focus a lot of my time and research on People Also Ask data because it’s the data we get from Google that updates the fastest. When something happens in the news, the PAA questions will change within an hour or two.

This information also tells us that SERP features are generated post-ranking. Google has their big churning system where they decide where webpages rank then, based on that and potentially some metrics they pump out afterwards, these features are generated. That’s how they can change so quickly. There’s no way to go into the guts of the ranking algorithm and have things change.

For years, we’ve had Googlers tell us that they don’t directly use clicks in ranking. I think they mean that Google does not use clicks as a quality indicator for individual websites. In my opinion, they’re doing two things with them. First, they’re building click models. For any given SERP, they have a predicted click-through score based on aggregated click-through data. It’s not used individually on any one particular site, so it ‘doesn’t matter’ if your site gets clicked on or not. I also think they use click data post-ranking.

The important thing for SEOs to understand (which fits in with intent shift) is that it’s query-based. Google is adjusting SERPs based on intent shifts around the query. It’s not to do with an individual site, how good it is, and where it ranks.

From Google’s point of view, your site might be a very attractive target for this keyword. However, Google might get data from the user that they’re after something else. It’s not that your site’s any better or worse, it’s that the user wants to aim at a different target. They’re not interested in what you’re saying anymore.

That’s missed by many SEOs. We get a blinkered view that a key phrase is a static target. We create the content then wonder why we don’t rank for it. It’s actually more to do with the fact that Google’s understanding of the query and the user has shifted, and you haven’t shifted along with them.”

How should an SEO strategy oscillate to stay on top of keyword and content targeting?

“One of the other details Google specifically mentioned, regarding systems like Instant Glue, was that old content gets more attention – simply because it’s there. It’s embedded, it’s growing, people know about it, and people have linked to it. Part of the function of Instant Glue was to artificially give more attention to new things.

There has long been a theory among different SEOs that there’s almost a trial period for new content that you publish. You publish something, it seems to rank for a little bit, then sometimes it drops out completely and other times it will stick, but maybe a bit lower down. The collective theory was that Google is giving your content a trial period and seeing how users like it.

We run a tool called AlsoAsked, and we released a timeline feature. You can see the history of any key phrase over time, the volatility, and how much the People Also Ask questions have changed around a topic. Of course, if you want to scrape it yourself, that’s absolutely possible. One of the main ways I track intent shift is by looking at how often those PAA questions change. It’s such a good canary in the coal mine.

To give some detail, you usually get 4 PAA questions attached to a query, and they are usually quite static. Branched off from these, for each 1 of those 4, there are another 4/5 questions. Therefore, there are around 100 questions, give or take, in any full tree.

The second-level questions tend to change fairly regularly because Google is experimenting, A/B testing, and using this as an induction loop to try and understand the queries better. Again, that’s something that became quite clear from the DOJ trial. A lot of Google’s magic is just seeing how users react to search terms and trying to iterate and improve on that.

Instead of using BERT to understand the nuance of a query, they’re chucking progressively slightly better stuff at the wall, seeing what sticks, and then chucking new stuff.

When we look at PAA data, we have our own volatility calculation. If one of the primary root questions changes, that’s a fairly big deal. If the second tier changes, that gives us a bit of extra volatility because some of them are quite static all the way down.

However, with the COVID example, the whole tree changed over 12-24 months, from the first root questions, as the collective knowledge of people progressed.

There are ways that you can do this programmatically, which are exciting. You can now monitor your content with Screaming Frog to see whether you have answered all the PAA questions. In version 20, they now have a custom JS functionality, and what you can achieve with it is absolutely fantastic. You can do crawls, and you can query APIs as you go.

I’ve been checking whether these PAA questions are changing against the content as we crawl it. We’re fetching the PAA data as we crawl and using ChatGPT to ask whether or not we answered those questions on the page. Fresh out of a crawl, we get a list of where the intent shift has changed and the new questions we need to answer.

If you’ve got a site with thousands of pieces of content, it can become unwieldy to manage and stay on top of what’s changing, but this is an automated process. You can get on with some other work and then get a report that tells you where the questions have significantly changed and where there are gaps in your content.

By updating that content, you’re lighting up Instant Glue by having new content, so Google’s going to give it a bit more attention. You’ve also matched it up with what people are searching for, so you’ve reduced the chances of people having to redefine queries because their questions have changed.

We’ve been having a lot of success with this. A lot of SEOs talk about updating content and putting out fresh content, and this is a really specific, targeted way to do that, so you get the most bang for your buck.”

Is there a tool that can predict likely search volume in the future?

“The crystal ball tool? I think that’s possible, but it’s human intelligence from people who know their sector and their industry.

If you asked me about SEO, we could make some good predictions about SearchGPT, generative AI search, and the kind of things we’re seeing right now that we know are going to be big. I’m not sure if any tool can encompass all of that yet, and predict those trends. If there is one, I would love to have access to it.”

If an SEO is struggling for time, what should they stop doing right now so they can spend more time doing what you suggest in 2025?

“A lot of SEOs spend too much time on metadata tasks.

Things like title tags are super important for SEO, but spending a lot of time on things like meta descriptions isn’t that helpful because the process is more query-based.

The stats say that somewhere between 50% and 80% of meta descriptions get rewritten by Google, and that’s done on a per-query basis. Google sees what the user has typed in, provides an appropriate page, and finds what it thinks that user wants from that page. They pluck that out and put it in a meta description. It’s in Google’s interest for you to be happy with the SERP.

If you are writing a meta description, you can only provide one static description of everything that’s on the page – and there might be 10, 20, 50, or 100 search terms for that page. Even in the best-case scenario, where you’ve managed to write something better than what Google would pull, it’s still only going to be shown 20-30% of the time. You’re immediately losing a huge chunk of value.

Also, Google’s recently announced they will be able to use the Open Graph titles they find in the HTML as the page title link within the SERP. We’ve seen Google rewriting titles in the SERP, and they commonly use the H1 of the page or the internal links.

That’s seemingly because SEOs are breaking away from what’s best for the user. They are writing title tags for search engines, but the H1 is what’s for the user. It feels as though Google is coaxing SEOs back and showing what they are after. They’re ignoring the title you’ve written and using what they think is appropriate. Google gets it wrong sometimes but, on average, that’s helping.

Again, things like title tags are important but Google’s rounding off those corners for us. You can spend more time providing value around your actual content. Spend more time on writing, having good meat, and making sure you understand what users are searching for and you’re making them happy – rather than polishing a meta description.”

Mark Williams-Cook is Director at Candour and Founder of AlsoAsked, and you can find him over at WithCandour.co.uk.