SEO in the Age of AI: What Still Works, What's Dead, and What's New

Search isn't dying. The strategy that ranked eighteen months ago is. If the entire SEO plan is still "rank for high-volume keywords," that's not the right game anymore.

Here's what's actually changed, and what to do about it.

What changed in the last 18 months

Three things stacked on top of each other and broke the old playbook.

Google AI Overviews now sit above the first organic result for roughly 30% of all queries, and the share keeps climbing. When the AI Overview answers the question directly, click-through to the underlying pages drops 30–60% depending on query type, with informational queries taking the worst hit.

ChatGPT and Perplexity have become meaningful traffic sources, not big yet, but growing fast. People are asking AI tools "best [category] brand for [need]" before they ever open Google. The brands cited in the answer get the consideration. The brands not cited don't even know they were in the running.

Google's helpful content updates and the spam updates of the last couple of years wiped out a lot of programmatic AI content that had been ranking thinly. Sites publishing generic AI fluff lost 60–80% of their organic traffic in a single update. Sites with real expertise, named authors, and real data held up. The bar moved.

What's dead

Generic "top 10 X" listicles with no original perspective, AI Overviews summarize these in two sentences and nobody clicks through. Keyword-stuffed content with no entity coverage, Google reads pages as concepts now, not keyword density. Programmatic AI content with no human review, recent spam and helpful-content updates specifically targeted this pattern. Backlinks as the primary lever, they still matter, but volume alone stopped being the move years ago. And optimizing purely for clicks on the results page, the visibility that matters now is citation inside the AI Overview and inside answer engines.

What still works, and works harder than ever

Real expertise, written by named humans. Google's quality raters look for the entity behind the content, and so do AI models deciding who to cite. A post by "Admin" on a generic site doesn't get cited. A post with a real byline, a real photo, and a real track record does. Show your face, cite your data, name your sources.

Definitive answers in the first hundred words. AI Overviews and Perplexity pull from the first chunk of a page that answers the question directly. Bury the answer under three paragraphs of context and you lose the citation. Lead with the answer, then expand underneath it.

Schema markup that actually matters now. Structured data used to be "nice to have." Now it's how AI models verify what a page is about. Article and Author schema on every post, Organization schema with sameAs links to social profiles, FAQPage on long-form posts (still wins featured snippets even with AI Overviews live), and BreadcrumbList site-wide. Worth checking for basic technical issues too, canonical tags accidentally pointing to the homepage instead of the actual page is a surprisingly common leak that quietly tanks crawl quality.

Brand searches as the new high-intent traffic. Someone Googling "[brand] vs [competitor]" or "[brand] reviews" converts at several times the rate of someone searching a generic category term. AI Overviews don't hurt brand searches, they often amplify them by surfacing a brand info card. The playbook shift for eCommerce isn't "rank for the generic category term," it's "make people search for the brand" through PR, social, podcast appearances, and third-party comparison content, then make sure the brand SERP itself converts.

Citation footprint over backlink volume. Brands mentioned across fifty or more relevant sources, podcasts, industry publications, founder interviews, niche forums, get cited in ChatGPT and Perplexity. Brands with five hundred random links from low-quality sites don't. Quality of mention beats quantity of link, and that's the real shift.

What's new and worth investing in

Generative Engine Optimization, the practice of making sure ChatGPT and Perplexity actually cite you. The early playbook: answer questions completely rather than partially, use clear declarative sentences (models struggle to extract hedged language), include current data and timestamps (models prefer recent sources), and aim to be cited alongside sites that already get cited, Wikipedia, Reuters, niche authority sites in the category.

Entity-based content architecture. Stop building content purely around keywords and build it around entities instead, the brand, service categories, founders, customer segments. Each entity gets a definitive hub page; supporting content links up to reinforce the parent entity. Google's Knowledge Graph runs on entities, and so do AI models, content organized the same way performs better across both.

Programmatic SEO, done properly. Programmatic content still works, but it has to be data-driven, not template-spun. Landing pages built from a real data source with original analysis behind each one perform very differently from thousands of generic pages spun from a single brief, the first compounds, the second risks a manual penalty.

What to do this quarter

Audit the top fifty pages by traffic and check how many got hit when AI Overviews launched in their query category, that's the urgent fix list. Rewrite intros to lead with the answer across every blog post, first hundred words. Add author bios with real credentials, real photos, real LinkedIn links. Run a schema audit, Article, Author, Organization, and Breadcrumb at minimum. And start tracking ChatGPT and Perplexity citations for the top twenty commercial queries in the category, free tools exist, and some paid ones are worth the cost.