Click-through rates from informational queries are down 35% year-over-year across the 47 sites we manage. AI overviews are eating the snippet, and on a meaningful chunk of queries the user gets their answer without ever leaving the SERP. If your SEO strategy was built for 2021, it's already broken.
What still works
Deep, opinionated, citation-worthy content. The AI overview has to cite someone, and increasingly that someone is the source with the most distinct point of view and the freshest data. Generic listicles regurgitating what's already indexed don't get cited — they get summarised away.
Original research wins disproportionately right now. Surveys, benchmarks, first-party data, even small studies of 200 people. We ran a 14-stack benchmark last month (see our other post) and it has been cited in AI overviews on 80+ queries. A normal blog post would have been cited on zero.
Product-led pages with real comparisons. 'X vs Y' pages where you actually compare the products in detail — pricing, feature matrices, screenshots, video walkthroughs — still rank and still convert. The AI overview rarely replaces these because the comparison is the value.
Structured data that helps the overview cite you instead of replace you. Mark up your authorship, your publication dates, your product specs, your FAQs. The overview is more likely to use you as a source when the data is machine-readable.
What doesn't work anymore
Thin SEO content built around keyword volume. Dead. The overview answers the query and the user moves on.
Listicles and 'ultimate guides' that just collect what's already on the first page. The model has read all of them and can synthesise a better answer in two seconds.
Programmatic SEO at scale with no editorial layer. Google's spam updates from late 2025 hammered this approach. We've seen sites lose 90% of their traffic in a single update.
The new playbook
Treat your blog like a magazine, not a content farm. Publish fewer posts, give each one a real author with a real opinion, invest in original data, and make every post worth being cited by an AI. We've moved most clients from 8 posts a month to 2, and traffic is up across the board.
Invest in entity SEO. Make sure your brand, your authors, and your products are well-defined entities in the knowledge graph. Wikipedia and Wikidata still matter. LinkedIn profiles for authors with real publication history still matter.
Build for the assistive search experience, not just the blue-link SERP. That means your content should be skimmable, factual, well-structured, and easy for a model to extract a clean answer from. If a model can't summarise your page in two sentences, neither can a human.
The metric that matters now
Stop obsessing over keyword rankings. Start tracking AI overview citations. Several tools (Profound, AthenaHQ, Otterly) now track when your brand appears in AI answers across ChatGPT, Perplexity, and Google AI mode. That number, not your position on page one, is the new north star.