SEO has been declared dead roughly once a year since 2005, and the people declaring it are usually selling the replacement.
What's true is narrower and more useful. The mechanics of search haven't changed much, but the payoff has. You can do everything right, rank first, and still get fewer visitors than you did two years ago, because 68% of US Google searches now end without a click to anywhere. The work still works. The scoreboard moved.
So here's the short version of what SEO is, what still matters, and what I'd do first.
What SEO actually is
Three questions, and every tactic you've ever heard of is answering one of them.
Can the machine find and read your site? That's technical SEO. Crawlability, indexation, site speed, mobile rendering, clean URLs, a sitemap that isn't lying. If a crawler can't get in or can't parse what it finds, nothing else you do matters.
Does your page answer the thing someone asked? That's content and intent. Not keywords stuffed into paragraphs, intent. Somebody typing "best CRM for a two-person team" wants a comparison, and somebody typing "how to import contacts into HubSpot" wants steps, and giving either of them the wrong format loses regardless of how well you wrote it.
Does anybody vouch for you? That's authority. Links, brand mentions, being cited by people who already have standing. This is the slow one, it's the one you can't shortcut, and it's the one that's grown more important, not less.
That's it. Everything else is detail.
What still matters in 2026
| Area | What to do | Honest weight |
|---|---|---|
| Technical health | Get indexed, stay fast, don't break rendering. Google's three speed thresholds: main content visible in under 2.5 seconds (LCP), responds to a tap in under 200ms (INP), and the layout doesn't jump while it loads (CLS under 0.1). | Price of entry, not a lever. Perfect scores won't save weak content. |
| Intent match | One page per real question. Match the format the searcher wants, not just the topic. | High. This is where most sites lose. |
| E-E-A-T (experience, expertise, authoritativeness, trust) | Real authorship, real credentials, real first-hand experience, an About page a human wrote. | High, and rising. |
| Internal linking | Point your pages at each other deliberately. Orphan pages are invisible pages. | Underrated, nearly free. |
| Structured data | Invisible code on the page that tells a machine what it's looking at. JSON-LD, covering Organization, Article, FAQ, and Product. | High and rising. It's how machines parse who's talking. |
| Being a primary source | Original data, original opinion, things only you can say. | The one that compounds. |
That last row is doing more work than it looks like. Google's March 2026 core update tilted visibility toward first-party and official sources and away from aggregators and content built primarily to rank, which is a polite way of saying the summarize-the-top-ten-results play is finally dead.
Two people worth listening to, who don't agree
I'd rather hand you a real disagreement than a consensus that isn't one. Here are two practitioners with long, public, checkable track records, pulling in different directions.
Lily Ray runs SEO strategy and research at Amsive and is probably the most rigorous public analyst of Google's quality guidelines working today. She's been reading core updates and E-E-A-T signals against real client data for years. Her position is the fundamentalist one, in the good sense: the underlying quality signals haven't materially changed, every core update just reshuffles their weights, and what lasts is being a real primary source rather than chasing whatever the algorithm rewarded last quarter. Do the boring work, and be the thing worth citing.
Kevin Indig advises growth teams and writes Growth Memo, and his framing is the structural one. He's documented what he calls the great decoupling, where rankings and traffic have come apart, so the position you hold no longer predicts the visits you get. He argues that search is turning from a ranked list into a definitive answer, that trust and brand are what decide who gets picked out of that answer, and that the SEO paying best now is increasingly product-led. Your actual product surfaces, feeds, and structured data doing the work blog posts used to do.
Where they differ: Ray's answer to a changing algorithm is to double down on quality and authority, because those are the constants. Indig's answer is that the constants still hold but the channel economics changed underneath them, so you need to change what you're optimizing for and how you measure it.
I lean Ray on execution and Indig on strategy, and I don't think that's a dodge. Ray is right that there's no trick and the boring work is the work. Indig is right that if you're still reporting SEO success as "we moved from position six to position three," you're measuring a thing that stopped mattering.
Rankings became a vanity metric while everyone was still putting them in the deck.
The first five things I'd do
In order, for a real business that hasn't done any of this:
- Set up Google Search Console. Free, takes ten minutes, and it's the one place you can see what Google actually thinks of your site. Most business owners I meet have never opened it, and that's not a knock on them. Nobody tells you it exists.
- Fix indexation before anything else. Search Console will tell you which pages Google can't or won't index. A page that isn't indexed can't rank, and this is boring and fixable and nobody does it.
- Write one great page per buying question. Five of them. The questions a prospect asks right before they'd contact you. Depth beats volume, always, and it beats it harder now.
- Add JSON-LD schema for Organization and Article, plus FAQ where it fits. An afternoon of work that pays into both classic search and AI answers.
- Put a real name and a real face on the content. Authorship, credentials, first-hand experience. This is the cheapest E-E-A-T signal there is and most sites skip it because it feels vain.
Notice what's not on that list. Keyword density, meta keywords, blog volume, backlink buying, and any tool with a monthly fee. You can get remarkably far before you need to spend money.
How this connects to AI search
The thing that surprises people is how much of this doubles. A site that's crawlable, fast, well-structured, clearly authored, and useful is a site both Google's index and an LLM's retriever can work with, because AI Overviews and AI Mode pull from the same index classic ranking uses.
But they aren't the same game. Only about 12% of the URLs AI tools cite overlap with Google's top-10 organic results, so being excellent at SEO doesn't automatically buy you a seat in the answer. That gap is what AEO is for, and I wrote the companion piece here: AEO: A Five-Minute Primer.
Do the SEO fundamentals first anyway. They're the foundation the other thing stands on, and unlike most of what gets sold as marketing, they keep paying after you stop.
If you'd rather hand the whole thing to someone, that's what I do.
Figures cited: SparkToro's 2026 zero-click study, Amsive's March 2026 core update analysis, and Conductor's 2026 AEO/GEO benchmarks. Positions attributed to Lily Ray (Amsive) and Kevin Indig (Growth Memo) are my summary of their published work, not quotes.
