Keyword research used to mean a spreadsheet, a search volume column, and a sort function. That approach still has some uses. It's just no longer the whole job, and businesses that treat it as the whole job are leaving gaps that AI assistants happily fill with a competitor's name instead of theirs.
Search volume stops being the main signal
A keyword tool will cheerfully tell you "boiler repair Colchester" gets searched eight hundred times a month and "my boiler keeps cutting out at night, is that dangerous" gets searched maybe thirty times. Old logic says chase the eight hundred every time. But that second, messier phrase is almost exactly how someone talks to ChatGPT late at night, worried and half-asleep. Low volume in a traditional keyword tool doesn't mean low value once an AI model is the one reading the question and picking a single answer to repeat back to a real person.
We still pull volume data for every client. We simply stopped treating it as the deciding factor it used to be five years ago.
A Colchester client pushed back hard on this at first. His instinct, reasonably enough after a decade in business, was to chase whatever number looked biggest on the report. It took showing him three months of AI citation data, side by side with the old volume report, before the penny dropped properly. He'd spent years paying for content targeting phrases that technically got searches but never once converted into a phone call, purely because the phrasing never matched how a worried customer actually spoke when they needed help in a hurry.
Questions matter more than phrases now
Traditional keyword research hunts for phrases to insert naturally into a page. AI-focused research hunts for genuine questions to answer completely and directly. The difference sounds small on paper. It isn't, in practice. "Conveyancing solicitor Chelmsford" is a phrase someone might type. "How long does conveyancing usually take if there's no chain" is a real question, and it's exactly the kind of thing a nervous homebuyer actually asks an AI assistant at eleven at night when they can't sleep.
- Pull genuine questions from customer emails, call transcripts, and WhatsApp enquiries you've already received
- Check "People also ask" boxes on Google, but don't stop your research there
- Look at the questions people ask in local Facebook groups specific to your town
- Note the exact wording used — "cheap" behaves differently to "affordable," which behaves differently again to "budget"
None of this requires expensive software. A notebook by the phone and five minutes after each call writing down the actual question asked beats most paid research tools we've tried, frankly. We ask every new client to keep exactly this kind of log for a fortnight before we touch a single page, because the raw material it produces is simply better than anything a keyword database can guess at from the outside.
Intent clusters replace flat keyword lists
Rather than one flat list of forty keywords, you end up grouping questions by what the person is actually trying to decide at that moment. Someone asking "is double glazing worth it" is still comparing options and gathering opinions. Someone asking "double glazing fitters near Rayleigh with good reviews" has already decided and is now choosing a specific supplier. These two readers need genuinely different pages, written in a noticeably different register, and lumping them together into one general "double glazing" page — which is exactly what a lot of older SEO advice still suggests doing — serves neither reader particularly well in the end.
This is honestly where a fair amount of SEO advice from five or six years ago starts to actively work against a business trying to compete today. We still see agencies elsewhere in the county selling clients a single consolidated page covering every stage of a buying decision, purely because it looks neater in a site map, and we think that's quietly doing those clients a disservice.
Competitor gap analysis looks different too, these days
Old-school competitor research compares which keywords a rival ranks for that you currently don't. The AI-focused version asks a different, sharper question entirely: when someone asks an assistant a question in your space, whose page actually gets cited as the source for the answer? We'll run twenty or thirty realistic customer questions through both ChatGPT and Perplexity for every client and log every single citation that comes back. It's considerably slower than pulling a standard keyword gap report. It's also far more honest about what's genuinely winning attention right now, rather than what won attention two years ago. For a Chelmsford furniture retailer we work with, this exercise showed a smaller rival's buying guide getting cited repeatedly, despite our client's page technically ranking higher on Google itself. The buying guide simply answered more of the actual questions people asked. We shared the finding with our client fairly bluntly, because sugar-coating it wouldn't have changed the underlying problem, and within a month their own team had drafted a competing guide covering the gaps we'd flagged.
A short word of caution on "AI keyword tools"
A fresh wave of new tools now promise to show you exactly what AI models are "ranking" for across your sector. Treat the output with a healthy dose of scepticism. Several we've tested produce confident, specific-looking numbers that simply don't match what we see when we actually go and run the exact same queries ourselves by hand. Use them for rough ideas and direction, not as gospel truth. The only genuinely reliable test, still, is typing the question into the assistant yourself and reading precisely what comes back on your own screen. We've had more than one client show us a glossy "AI visibility score" report from a third-party tool claiming they ranked well, only for the actual assistant to return nothing about them at all when we tested it live in front of them.
Writing for the research you've gathered, not just doing it
Research that never actually changes a single page is wasted effort, full stop. Once you've built your list of real, specific questions, the content needs to answer them in the first two or three sentences of a section, not bury the actual answer under three paragraphs of polite throat-clearing first. This one habit is among the most reliable levers we use in client content work across the county, and it costs nothing beyond genuine discipline in how a page gets drafted and edited.
Local nuance still counts for a great deal
Don't lose specific, Essex-flavoured detail while chasing generic, AI-friendly phrasing elsewhere. "Parking near our Maldon showroom is free after 6pm on weekdays" answers a real, slightly anxious local worry better than any amount of clever keyword insertion ever could manage, and it's exactly the kind of concrete, specific line an AI assistant will happily quote, because it genuinely sounds like a real answer rather than polished marketing copy written to tick a box.
We built exactly this kind of hyper-local research into a wider piece of keyword strategy work for a Billericay estate agent earlier this year, pulling questions directly from branch viewing feedback rather than guessing what buyers in that specific postcode actually cared about.
The Bottom Line
Keyword research for AI answers isn't a wholesale rejection of everything that came before it — volume data, competitor checks, and search trend tracking still all matter, just not as much as they once did on their own. What's genuinely changed is the priority order. Questions now outrank raw phrases, intent clusters replace flat lists, and the true test of your research isn't a tidy spreadsheet any more, it's whether an assistant reads your page straight back to someone as the answer. Start with the messy, specific questions your actual customers ask, and build everything else outward from there. It takes longer than copying a keyword list from a tool. It also tends to hold up far better over time, since real questions change more slowly than search fashions do.
Looking for SEO agency in Essex?
Let's talk about what your business actually needs
Digital Impact Ventures provides SEO, AI visibility and digital marketing support for businesses across Colchester and Essex. Get in touch and we will talk through where your business is now and where it needs to go.