- Backlinko's core on-page principles still move rankings today
- AI Overviews can eliminate clicks even at rank one
- Pew Research found AI summaries roughly halved click-through rates
- Structure pages so AI systems can extract clean direct quotes
- Track AI citations alongside rankings and iterate on real data
Last spring I audited a B2B SaaS client that owned the number-one spot for a fat commercial keyword — and still lost roughly a third of its clicks in a single quarter without slipping a single position. The culprit wasn’t a competitor. It was an AI Overview parked above result one, answering the query before anyone bothered to scroll. That was the moment Backlinko’s on-page SEO playbook stopped being enough for me. The framework wasn’t wrong. It was built for a search engine that hands back ten blue links, not one that increasingly writes the answer itself. So here’s what I want to walk you through: what still holds up, what quietly broke, and how a data-driven SEO approach patches the gap between ranking first and actually being the answer.
What Backlinko’s On-Page SEO Framework Actually Says
Backlinko’s guide codified on-page SEO into a repeatable, correlation-backed checklist that a generation of marketers treated as gospel. Brian Dean’s on-page SEO guide earned its authority the hard way — by pairing plain-English tactics with large-scale ranking-correlation studies at a time when most SEO advice was pure folklore. That’s why it became the default reference. It gave you concrete moves you could execute on a Tuesday afternoon.
Strip it to the studs and the framework says a handful of durable things. Here’s how I summarize its core tenets when I brief a new writer.
This table maps each classic tenet to the job it was designed to do.
| On-Page Element | What Backlinko Recommends | The Job It Does |
|---|---|---|
| Title tag & URL | Front-load the target keyword | Relevance signal to the crawler |
| H1 & headings | Include primary and variant keywords | Topic clarity for indexing |
| Content depth | Cover the topic comprehensively; longer often correlates with rankings | Perceived thoroughness |
| Internal linking | Link to and from relevant pages | Authority flow and crawl paths |
| Multimedia | Add images, video, visuals | Engagement and dwell time |
| UX & dwell time | Fast, readable, low bounce | Behavioral quality signals |
Notice the through-line. Every one of these is a ranking-factor optimization — a bet about what correlates with a higher position in a list of links. That’s the framework’s DNA, and also its ceiling. It was never built to model search intent in the way engines now do, and it certainly wasn’t built for answer engines that quote you instead of linking you. It’s a correlation map of a world that’s being redrawn.
Where the Classic On-Page SEO Checklist Still Holds Up
The fundamentals Backlinko popularized remain necessary — they’re just no longer sufficient on their own. I’ve never once regretted matching a page tightly to search intent or shipping a clean heading structure. Those moves still move rankings, and they now double as the raw material AI systems extract from.
Here’s what I still treat as non-negotiable, straight from the classic playbook:
- Search-intent match — the single highest-leverage on-page decision, then and now.
- Clear structural hierarchy — descriptive H2s and H3s a machine can parse in one pass.
- Genuine content depth — not word count for its own sake, but full coverage of the question.
- Technical hygiene — indexable, fast, crawlable, with sane internal links.
Here’s the honest split between what still carries a page and where it now runs out of road.
| Classic Tenet | Still Essential? | What It No Longer Covers |
|---|---|---|
| Intent match | Yes | Answer-ready phrasing for extraction |
| Content depth | Yes | Entity coverage and cited evidence |
| Heading structure | Yes | Standalone answers a model can quote |
| Technical hygiene | Yes | Visibility inside AI answer surfaces |
I watched intent-matching alone rescue a stalled page last year. We didn’t add a word of length — we re-sequenced the H2s to answer the actual query first, and it climbed from page three to the bottom of page one in about six weeks. But page one only got us halfway to the goal now, which is a different problem than it used to be. If you want the deeper version of how we work these signals, I laid it out in our breakdown of a performance-data-led SEO strategy.
What’s Changed in On-Page SEO Since Backlinko’s Guide?
Search stopped being a list you rank in and became an answer you either get quoted in or get left out of. Three shifts drove that, and each one bends on-page priorities in a direction the classic checklist doesn’t cover.
First, AI Overviews and the broader generative experience changed what a top ranking is worth. When Google synthesizes an answer at the top of the page, the click you earned by ranking can evaporate. Pew Research Center found in 2025 that users were roughly half as likely to click a traditional link when an AI summary sat at the top of the results. That’s the zero-click reality, and it punishes pages built only to rank.
Second, the engine reads for meaning now, not string matches. Google’s helpful content and passage-indexing systems reward semantic and entity-level understanding over exact-match keyword placement. Their own people-first content guidance and E-E-A-T signals reward demonstrated experience and expertise, not keyword density. Peter, our founder, unpacked the practical fallout in his take on Google’s helpful content update.
Skip the spreadsheets. Start Sage SEO free.
See how AI content ops transform your agency workflow in minutes.
Third — and this is the one the old checklist has no vocabulary for — being citation-worthy now matters as much as being readable. An answer engine has to be able to lift a clean, self-contained sentence out of your page. If your best insight is buried in paragraph nine or trapped inside a clever metaphor, it doesn’t get quoted.
This table shows how each shift reprioritizes the work.
| Shift | Old On-Page Priority | New On-Page Priority |
|---|---|---|
| AI Overviews & zero-click | Rank #1 for the click | Get extracted into the answer |
| Semantic / entity search | Keyword placement | Topic and entity coverage |
| E-E-A-T & helpful content | Comprehensiveness signals | Demonstrated experience & trust |
| Answer-engine citation | Readable for humans | Structured for extraction |
How Sage SEO’s Data-Driven Approach Builds On the Framework
We keep every fundamental Backlinko taught and wrap it in a measurement loop, so on-page elements become hypotheses we validate instead of rules we obey. That’s the core difference. A static checklist tells you what worked in aggregate across thousands of sites. A feedback loop tells you what’s working on your page, this month, for this query.
Here’s the sequence I run, and it’s designed to serve both a Google ranking and an AI citation at the same time:
- Start from a real search signal. Every page begins with observed demand — an impression spike, a rising query, a gap in the cluster — not a hunch about a keyword.
- Draft for extraction, not just for reading. One-sentence direct answers under question headings, scannable structure, and clean tables that a language model can lift verbatim.
- Interlink into the cluster. Every new page connects to and from its topical neighbors so authority compounds instead of fragmenting across orphan URLs.
- Measure, then rewrite. We watch how the page performs in classic rankings and in AI answer surfaces, then treat the underperformers as experiments to iterate.
That last step is where I see the most surprised faces. Marketers still ship on-page as a one-time optimization — publish, tick the box, walk away. I did it that way for years and left revenue on the table. The pages that win now are the ones you revisit when the data tells you the answer engines skipped you. Tracking whether ChatGPT, Perplexity, and AI Overviews actually cite a page is its own discipline, which is why we built cross-engine citation tracking to watch it directly rather than guess.
The structural point matters too. Crawlers index your page; language models parse it. Those aren’t the same act. A model rewards a page that states its claim cleanly, attributes its evidence, and organizes facts into rows it can quote — which is exactly the pre-publish work I outline in our guide to prepping a post for AI Overviews and ChatGPT citations. Backlinko optimized for the crawler. We optimize for both, and we let performance data settle the arguments.
Backlinko vs. Sage SEO: A Side-by-Side Look
This isn’t a rejection of Backlinko — it’s the same fundamentals run as a living process instead of a fixed formula. When I compare the two side by side for clients weighing frameworks, the contrast lands in four dimensions.
| Dimension | Backlinko’s Classic Framework | Sage SEO’s Data-Driven Approach |
|---|---|---|
| Keyword strategy | Keyword placement in title, URL, H1 | Intent and entity coverage, extraction-ready phrasing |
| Content structure | Readable, comprehensive, long-form | Structured for both crawlers and answer engines |
| Measurement | Correlation studies across many sites | Your page’s own performance, tracked and iterated |
| AI-search adaptability | Built pre-AI-Overviews; not designed for citation | Built to be quoted, cited, and re-measured |
Read those columns honestly and you’ll see the left one isn’t obsolete. It’s the foundation. What’s changed is that a foundation is no longer the whole house. The classic guide gives you a strong, static blueprint; a modern approach treats that blueprint as version one and keeps shipping revisions against real feedback. If you’re refreshing an older strategy doc, our view on building a durable, connected SEO strategy walks through how the on-page and off-page pieces should reinforce each other rather than compete for attention.
Practical Takeaways: What to Apply Today
Audit your best pages against both frameworks at once — keep what still ranks, add what now gets you cited. This is the checklist I actually run, and none of it requires ripping out your existing on-page work.
- Lead every section with a direct answer. Make the first sentence under each heading a complete, standalone claim an engine can quote.
- Keep the classic fundamentals. Intent match, clean headings, real depth, technical hygiene — none of that expired.
- Structure facts into tables. Comparisons and data belong in rows, not buried in prose, because that’s what gets extracted.
- Cite real sources with links. Named studies and authoritative references build the E-E-A-T that models and raters reward.
- Re-measure and rewrite. Treat on-page as an ongoing loop; revisit pages the answer engines ignore.
Run one existing page through that list this week. I’d bet the fundamentals are already solid and the extraction layer is what’s missing — that’s the pattern I see in nine audits out of ten.
The Blueprint Held — The House Is Being Rebuilt
Backlinko didn’t get on-page SEO wrong; it got it right for the search engine of its era. The tenets it hardened into a checklist are still the load-bearing walls, and I’d never advise anyone to tear them out. But the ground shifted underneath them. Ranking first and being the answer are two different jobs now, and only one of them fills the pipeline. The framework isn’t dead — it’s being rebuilt in real time, page by page, against data the original studies never had access to. Keep the fundamentals. Add the loop. Then go check whether the machines are actually quoting you.