GEO: The Search Channel Nobody Budgeted For

In January, AI assistant referrals were a rounding error in every analytics property we manage. By June they were a line item worth discussing in client meetings. That is a fast curve, and it is why generative engine optimization, GEO, has gone from conference-panel filler to an actual workstream at our studio.

Let me try to cut through the noise, because there is a lot of it. Half the industry is selling GEO as a replacement religion for SEO, and the other half is dismissing it as a rebrand of things we already do. Both are wrong in useful ways.

What is actually different

Traditional search ranks documents. Generative engines synthesize answers, and your brand either exists inside that synthesis or it does not. That changes the optimization target in three concrete ways.

First, citations beat rankings. When an assistant answers a category question, it draws on a handful of sources and sometimes names them. Being one of those sources matters more than being position three on a results page nobody visits. The sources these systems trust skew toward the genuinely authoritative: original research, detailed comparisons, content with named authors and verifiable claims.

Second, entities beat keywords. These systems reason about brands as entities with attributes. If the model's picture of you is thin or outdated, you lose recommendations you never knew were happening. Consistent, factual, structured information about who you are and what you sell, everywhere it can be crawled, is now table stakes.

Third, third-party corroboration matters more than ever. Assistants weight what others say about you heavily, because that is how they estimate trust. Reviews, press, community discussion, comparison posts on sites you do not control. Classic PR, suddenly measurable in a new place.

What we are actually doing

Our GEO playbook for clients right now is deliberately unglamorous:

• Publish original data. We helped Northwind turn their internal category data into a quarterly report, and it now gets cited in generated answers about the space. One asset, compounding returns.
• Restructure key pages so a machine can lift clean answers: clear claims up top, specifics, named authors, dates, schema markup done properly.
• Audit what assistants currently say about the brand. We run a standing battery of category prompts monthly and track mentions, accuracy, and sentiment. Crude, but it is the closest thing to rank tracking this channel has.
• Fix factual inconsistencies across the web. Old pricing, dead product names, and contradictory descriptions actively poison the entity.

Notice what is not on the list: exotic tricks, invisible text, prompt-injection nonsense. Some of that stuff circulates in growth forums and it will end exactly the way early black-hat SEO ended.

Do not torch your SEO budget

Here is the part the hype merchants skip. Generative engines learned to trust the same signals search engines trust, because they were largely trained on the same web. The site with fast pages, clean structure, real expertise, and honest link equity wins in both systems. GEO is less a new discipline than a new distribution surface for the old discipline done well.

Where budgets should shift is measurement. Zero-click answers mean your best outcomes are increasingly invisible in referral data, so we are leaning harder on branded search volume, direct traffic trends, and post-purchase surveys asking how customers first heard of the brand. Imperfect instruments, but honest ones.

My prediction for the back half of 2025: GEO consultancies will multiply, most will be repackaged SEO, and the winners will be brands that simply became the most citable source in their category. Be citable. Everything else is commentary.