Guides
How to do SEO in 2026: the two-surface playbook
How to do SEO in 2026 is a different question than it was even two years ago, because the surface split in two. Google still ranks pages, and those fundamentals still compound. But a growing share of decisions now form inside answers, AI Overviews on the results page and assistants like ChatGPT, Gemini, Claude and Perplexity before it, where there are no positions at all. Here is the playbook that covers both, and what to stop doing.
Daniel Arons · Co-founder & CEO, Aethon AI
Eight years building digital marketing programs across SaaS, financial services, and consumer brands · Updated July 2026
The fundamentals that still compound
Technical health: indexable, fast, stable pages remain the entry ticket to both surfaces. Intent-matched content: one page per real buyer question, answering directly in the first lines, structured so a human can scan it and a model can lift it. Genuine authority: sources, data and expertise no competitor can paste. And measurement in business terms: organic pipeline, not rankings screenshots.
What to stop doing: publishing commodity posts that restate the top ten results, chasing keyword variants with near-duplicate pages, and treating schema as a growth hack rather than plumbing. Google said the quiet part out loud this year, there is no technical trick that substitutes for being the best answer, our full take is in the GEO is still SEO debate.
The 2026 addition: the answer layer
Two habits separate modern programs. First, write for questions and moments, not keywords: buyers confide situations to assistants, we lost a renewal over SOC 2, my jeans do not fit, and get recommendations before any query exists. Your content plan should cover those situations, which no keyword tool will show you, the method is in life moments marketing.
Second, measure answers like you measure rankings: your money questions through all four assistants monthly, share of voice per assistant, citations logged so every loss comes with a repair address. The metric stack is in AI share of voice, and the difference between the surfaces in AI Overviews vs Google Search.
A quarter-one operating rhythm
Month one: fix the foundation in the order from how to fix SEO, and baseline the answer layer with a free AI visibility audit. Month two: publish direct answers for your ten highest value unclaimed questions and correct the third party sources assistants cite. Month three: re-test everything, keep screenshots, and report both surfaces in one document: rankings and clicks next to share of answers and AI referred pipeline. Teams that run this rhythm stop arguing about whether SEO is dead and start compounding on both surfaces at once.