“God, this makes me so glad that I never learned SEO because those rules sound awful. This GEO stuff makes so much more sense to me.” – One of my Signal members said that during a live training session, and I laughed and understood completely — some of those old school SEO rules are kind of a mess!
But then I had to give the caveat: “But depending on what platform you’re optimizing for, some of those old keyword-matching techniques still apply.”
She groaned… and I cackled.
Content platforms are on a spectrum between keywords and synthesis
Just like there’s no one-sized-fits-all marketing strategy for your business, there’s also no “one content strategy fits all platforms.” Spending on what type of content you create (written, audio, video, images) and where your audience is looking for information, the platforms you choose may differ… and your search optimization strategy will change with it.
These platforms aren’t just “old school SEO” or “new school GEO.” It’s a spectrum, and almost every platform you’re already using sits somewhere in the middle of it, not at either end. Some are still deeply keyword-dependent. Some have layered real machine learning on top of a keyword foundation. Some have moved almost entirely toward behavior and semantics. And a few are running full synthesis, blending sources into a generated answer.
So I’m going to share where each of these platforms line up now, in summer 2026 — but remember, platform algorithms shift constantly, so treat this as a snapshot, not gospel or prophecy. But the ordering itself — and the reasoning behind it — is useful for figuring out how to write for wherever you’re actually publishing.
Still largely classic SEO keyword matching: Apple Podcasts & Pinterest
Apple Podcasts is the most old-school platform in regular rotation for most of us, still running on 2015-era SEO. Show title, episode titles, and now auto-transcripts get weighed against the literal words someone typed into search. There’s no dedicated keywords metadata field,so the actual lever is exact phrasing in your titles and descriptions.
Pinterest is close to old-school SEO, though it’s shifting. Titles, descriptions, and board names still function as literal keyword-match fields. But Pinterest has layered an internal “annotation” system on top — invisible tags it assigns based on both the text and the image content — plus entity recognition for topical authority. So: keyword-matching with a real semantic layer bolted on, but not fully synthesized the way a generative system is.
Hybrid: classic SEO keyword layer with machine learning layer stacked on top
Amazon runs a category-and-keyword backend that’s still very classic in structure, but also tracking user behavior signals and using those to influence their recommendations based on things like what books are commonly purchased or downloaded. (See my full explanation over at Happily Ever Indie of the eight factors of the Amazon algorithm.)
Amazon’s system is neither pure keyword-match nor full synthesis — it’s a mix of both.
Mostly semantic and behavioral now, keyword-matching secondary
YouTube has shifted the furthest of any platform that I’ve been teaching for more than five years. As of 2026, YouTube’s own documentation and outside analysis both describe it prioritizing semantic and topical relevance, plus viewer-satisfaction signals — completion rate, click-through, whether the person’s actual search intent got met — over literal keyword density in your title or description. Exact-match keyword stuffing can now actively work against you; YouTube’s own language discourages it outright.
Full synthesis platforms
ChatGPT, Gemini, and Perplexity blend multiple sources into one generated answer. There’s still keyword-matching happening underneath, at the retrieval step — the system has to find candidate sources before it can synthesize from them — but the surface behavior a user actually experiences is synthesis, not a list of matches.
And Google’s AI Overviews also use full synthesis for their answers (see this full post about how Google’s search engine results page is a mix of synthesis, extraction, and matching).
So how are we supposed to optimize our content when the rules are different?
It’s not “podcasts and images are old-school while videos and search engines are new-school.” Every platform sits at its own point on the same spectrum, and the general rule is: platforms with simpler, older backend infrastructure (Apple Podcasts) still reward literal keyword matching, platforms that have invested heavily in machine learning (YouTube, increasingly Amazon) have moved toward behavioral and semantic signals, and full generative engines sit furthest along that same line.
One thing holds steady across the entire spectrum, regardless of where a platform sits: fluency — writing in the genre-accurate language of wherever you’re publishing — pays off everywhere. It just gets weighted differently depending on how far along the spectrum that particular platform has moved.
- Apple Podcasts wants your exact phrase in the title.
- YouTube wants your content to actually satisfy the search intent behind the phrase.
- ChatGPT wants your point stated clearly enough to survive being restated in someone else’s words.
Even though each of these platforms use different mechanisms, they’re dependent upon the same underlying discipline: Say the specific, accurate thing, in the words your clients would use.
At least that hasn’t changed.
