Let’s say you supply industrial dust extraction systems. A plant manager asks an AI search tool whether you can retrofit a system during a weekend shutdown. It recommends two competitors and describes you as a supplier for new installations, citing an old distributor page. You’ve been doing retrofits for years. The plant manager now has two other firms to call.
Meanwhile, your SEO report shows the product page ranking well and your GEO report records a brand mention. Both are accurate. Neither tells you, on its own, that a prospective customer got the wrong idea about what you do.
This is what irritates me about the SEO versus GEO fight. A change in how customers research a purchase has become a dispute over budgets and job titles. The business can pay for two sets of reporting and still leave the same questions unanswered.
Google is talking about Google
In its guide to generative AI in Search, Google says its AI features use its core ranking and quality systems to retrieve relevant pages. It considers optimising for those features part of SEO. Fair enough. It’s describing how its own Search features work. That last bit tends to disappear when people quote Google to settle the entire debate.
Microsoft Advertising’s guide also stresses the importance of SEO, while giving more weight to how information is phrased and structured for AI systems. There’s a worthwhile argument here about what changes when a system compares suppliers and gives the buyer an answer. “SEO is dead” and “GEO is just SEO” save us the bother of having it.
SEO can help you see what buyers search for, which pages they find and whether the evidence they need is easy to reach. Generative engine optimisation, or GEO, gives you another set of questions. What did the system say about the business? Where did it get that information? Did it recommend you for the right job, or use your website as a source and send the buyer to somebody else?
Put those questions together and the work gets easier to prioritise. If the retrofit case study is buried in an unlinked PDF, make it easier to find. If a distributor is still describing your old offer, get the description corrected. If nobody has published what it takes to complete an installation over a weekend, talk to the people who’ve done it. I’d be a bit annoyed if two agencies billed me separately to discover the same missing case study.
What are we paying to fix?
A publisher has a good reason to worry about losing visits even if its articles are being cited more often. A visit can earn advertising revenue or lead to a subscription. In its study of 900 US adults’ browsing behaviour, Pew found that people clicked a traditional result on 8% of Google visits with an AI summary, compared with 15% without one. That’s an observation from March 2025, across different mixes of queries. For a business that earns money from visits, it’s a good reason to investigate what’s happening to its own traffic.
A retailer might have a different problem. Adobe’s analysis of US retail traffic found that AI referrals generated more revenue per visit than other traffic in September 2026. Before shifting a budget, I’d want to know how many of those visits the retailer gets and what they contribute to total sales. Our industrial supplier has a different question again. How often is it being ruled out for work it could do, before anyone reaches the website or speaks to sales?
Telling all three businesses to put a fixed percentage of their SEO budget into GEO is a remarkably confident way of avoiding the actual decision. I’ve written before about what happens when GEO wins budget at the expense of the work it depends on. Before approving a new line of spend, ask what it will buy. A tool that flags a recurring misrepresentation may be worth paying for. So might an afternoon with the installation team to document what the supplier can actually do. Those solve different problems.
Give sales and support something back
At BrightonSEO I talked about using search data as market intelligence. Look for patterns in the searches. What are buyers trying to solve, what’s making them hesitate and what are they comparing? A query marked “commercial” tells sales very little. A cluster of searches about retrofitting an existing line, keeping part of the plant running and finishing over a weekend tells them what the buyer is worried about. How much production are we going to lose?
Sales can tell us what comes up once people get in touch. Search gives us a way to investigate the questions people try to answer before they make that call. In 6sense’s 2026 B2B buyer research, 81% reported using an LLM during an evaluation and 26% reported using one to discover vendors. That’s self-reported behaviour in substantial B2B purchases. It suggests that buyers can already know who you are and still use AI to check whether your equipment fits, compare you with another supplier or decide whether you’re worth contacting.
The evidence needs to stay attached to where it came from. Search Console gives you a partial view of searches where your site appeared. Bing’s grounding queries show phrases associated with retrieval and citations. If your agency writes a prompt and runs it twenty times, it has run twenty tests for the same thing, but not shown that this is relevant. To find out whether customers actually ask that question, is a different task, look for it in observed searches, enquiries, call notes and support tickets.
Then give sales and support something they can use. Sales can ask about the shutdown window on the first call and share a case study showing how long the line was actually down. Support can document what customers need to have ready before the crew arrives. If buyers repeatedly need two days and the installation team needs three, take that to operations. Maybe more preparation can happen off-site. A new page about fast installations would be getting a bit ahead of ourselves.
Give somebody the authority to fix it
Dedicated GEO expertise makes sense. Tracing a claim across answers and external sources takes different skills from diagnosing an indexing problem. The trouble starts when each team is told to improve its own numbers and everyone assumes somebody else has the distributor’s phone number. In Muck Rack’s 2026 survey of 971 PR professionals, mostly in the US, 29% said nobody owned GEO in their organisation.
Give somebody responsibility for deciding what gets fixed first and the authority to involve the people who can fix it. Keep search visits, citations and qualified enquiries separate in the reporting. Use them together when deciding where to spend time and money. In the retrofit example, the distributor description, the case study and the answer sales gives about the shutdown window all need to agree with what the installation team can deliver.
At the next review, I’d ask both teams to try what the plant manager tried. Can they find out whether we can fit the system over the weekend and have the line running by Monday? If they still have to ring the engineer, I’d spend the next bit of the budget getting that explanation onto the website.
What should your search budget achieve?
If you’re weighing up where you’re likely to get a better return from SEO and GEO, I can help you decide what’s worth funding first, what can wait and how you’ll know whether the investment paid off.