Your buyers have stopped opening ten tabs. They ask a question, read one synthesized answer, and your company is either named in it or it isn't.
Generative Engine Optimization is the work of making your content the material an AI system uses when it writes that answer. Most of what gets sold under the term is old SEO with a new label, or software that measures a problem it can't fix.
This is written for B2B service firms with real expertise and no marketing department to speak of. It covers what has changed, what hasn't, and where to spend a limited budget.
What Changed
Search ranked pages. Generative search assembles answers from pieces of pages. That shift has three practical effects.
The unit of competition is the passage. A 2,000-word article with one sharp, well-supported paragraph beats a 2,000-word article that is consistently fine, because the model only needs the paragraph.
The outcome is a mention rather than a visit. Being named in the answer as the firm that made the useful point is the win. Traffic may follow later.
Hedging costs you the citation. If your page says "it depends on many factors," the model will use a source that commits. Nobody has ever been quoted for saying it depends.
What Didn't Change
The sources AI systems cite are, in the main, the sources that were already good by any older standard: clear structure, the answer in the first two sentences, headings that match how people phrase questions, a named author with checkable credentials, an original observation instead of restated consensus, pages that load and can be crawled.
This is the SEO advice competent people were giving in 2015. What's gone is the margin for ignoring it. In 2019 you could rank a mediocre page with enough links and patience. In 2026 a mediocre passage has a thousand alternatives sitting next to it, and the model has no reason to be generous.
The Test
Every paragraph you publish can be checked against one question:
If a knowledgeable, sceptical person were writing an answer on this topic and needed a source, would they reach for this sentence?
Most B2B content fails, for reasons that repeat across firms.
It describes when it should assert. "Our approach to positioning is holistic and tailored" contains no claim. "Most premium service firms lose the positioning argument in the first line of their homepage, because they lead with what they do instead of who it's for" contains a claim a reader can disagree with. Only the second is worth quoting.
Nobody wrote it. Content credited to "The Team," or to no one, carries no authority. Readers and models weight the same signal: a named person with a track record who is willing to put a claim under their own name.
It says what the other fourteen articles say. If your piece on cloud migration makes the same five points as everything ranking above it, there's no reason to pick yours. Consensus is filler, and filler doesn't get referenced.
The point is on page two. Three paragraphs of context, then the claim. The reader and the crawler have both left by then.
What to Do, in Order
For a firm with expertise and limited marketing capacity, this is the sequence. Do it in this order, one step at a time.
1. Audit What You Already Have
Run the existing site and resource pages through the test above. Most firms find two or three passages worth keeping and a large amount of material that reads well and commits to nothing. That material isn't harmless. Every thin page is evidence that your domain publishes filler. Cut it or rewrite it before adding anything.
2. Decide What You Want to Be the Source On
Pick a specific question, or a small cluster of them, where your firm has a defensible answer that comes from having done the work and that the generic answer gets wrong. Two or three of these is a strategy. Twenty is a content calendar with nothing behind it.
The internal question to ask: when a buyer puts this topic to an AI system, what do we know that the standard answer will miss?
3. Write the Answer First
Open with the claim. Skip the context, skip the "in today's rapidly changing landscape." State the thing as plainly as you can defend it, then give the reasoning and the evidence. Structure the rest under headings phrased as the questions a buyer would actually type.
People trained to build an argument gradually find this uncomfortable. It's still the only structure that works with a system that reads the first hundred words and decides.
4. Put a Name on It
A real author, with a real bio linked from the page, giving the specific experience that makes this person credible on this specific topic. Founder and executive content is the authority mechanism here, whatever it looks like from the inside.
5. Make Sure a Machine Can Read It
Clean HTML, sensible heading levels, structured data where it applies, no critical content locked in images or scripts, fast pages, no crawl blocks left over from the last redesign. None of this earns a citation by itself. Any of it can cost you one.
6. Publish Where the Models Look
LinkedIn articles, contributed pieces in industry publications, and answers in the forums where your buyers ask questions all feed the same systems. One well-argued point published in three places under the same name does more than three points published once each on a page nobody links to.
What to Stop Doing
Stop producing volume. A weekly generic article was a weak tactic under classic SEO. Under generative search it works against you, for the reason in step one.
Stop buying GEO dashboards before you have something to measure. A tool that reports how often ChatGPT mentions you is useful once you've written the thing worth mentioning. Before that it is a subscription that reports zero.
Stop publishing AI drafts as finished content. The models answering your buyers were trained on the same material your draft was assembled from. Feeding it back adds nothing and gets cited by nothing. Use the tools for structure and speed and supply the judgment yourself.
Why This Suits Firms Without a Marketing Department
GEO rewards the thing a B2B service firm with real expertise already has: people who know where the standard answer is wrong. It penalises the thing those firms usually outsource: a steady supply of competent, anonymous, agreeable content.
The work is to identify the handful of things your senior people know that are worth being the source on, get those written down with a name attached, and set up a repeatable process for doing it again next quarter.
A junior can run that process once it exists. A junior can't invent it. The senior judgment goes in at the start, into deciding what the firm should be cited for. After that the system runs.
This is how we work. We build the content engine, document it, teach your people to operate it, and step back on a date agreed at the start. A firm that owns its own process doesn't need an agency to stay visible. It needs one, occasionally, to look at what comes next.
Want to build this process for your own team? Get in touch and we'll show you what it looks like for your firm.