I have had the same on-page SEO checklist since 2015 – a simple Excel sheet with columns for the on-page factor, what to check, and a comment. Favicon. Title tag. Meta description. Canonical tags. H1 to H6 tag sequence. It has been on every client audit I have run for years, across international campaigns for clients in the US, UK, Canada and Australia.
A few weeks back I opened it again and something felt off. Not wrong, exactly. Dated.
One row still said to check “Google Webmasters.” Another told me to verify keyword frequency, density, prominence and proximity – language from an era when you optimized for a crawler, not for a person asking a question. There was a row for mirror sites. There was a rule that said H2 tags should appear a maximum of two times, H3 a maximum of four, and so on, like Google was counting.
None of this is a small tweak. Search itself has moved – Core Web Vitals, E-E-A-T, and now AI Overviews and answer engines like ChatGPT and Perplexity pulling from the same pages we optimize for Google. A checklist that doesn’t account for any of that isn’t a checklist anymore. It’s a museum piece.
So I sat down and rebuilt it, row by row. Here is what I kept, what I threw out, and what I added.
What still holds up
A good chunk of the original list survived without a single edit. HTTPS. XML sitemaps. Robots.txt. Canonical tags for domain and page-level duplication. Custom 404 pages. Internal linking. Broken link checks. Image alt text.
This is the part worth saying out loud: the fundamentals of technical SEO have not changed in a decade. What changed is the language we describe them in, the tools we use to check them, and everything that got added on top.
What quietly broke
A few rows were still technically true but pointed at the wrong tools or the wrong names.
“Google Webmasters” became Google Search Console back in 2015. Universal Analytics, which the sheet assumed, was sunset in 2023 – it’s GA4 now. Google’s old Structured Data Testing Tool is retired; you validate schema with the Rich Results Test today. “Website speed” as a generic line item doesn’t mean anything anymore – it’s Core Web Vitals, and since March 2024 the metric that matters is INP, not FID.
Other rows were actively bad advice by today’s standards. The fixed H-tag repeat counts (H1 once, H2 twice, H3 four times, doubling down each level) is a rule from an era of literal keyword placement, and following it today reads as keyword stuffing to a Helpful Content-era algorithm. Same with checking “keyword frequency, density, prominence and proximity” as a standalone line – that’s optimizing for a 2012 crawler, not a 2026 search intent match.
And a couple of rows, like the mirror sites check, weren’t wrong so much as rare. Duplicate-site setups used to be a real risk in the load-balancing era of web hosting; today’s CDN and hosting infrastructure makes it mostly a non-issue, though the underlying risk (duplicate content across domains) is still worth folding into a general duplicate-content check.
What I added for the AI-search era
This is the bigger shift. The old sheet had no concept of a page being read by anything other than Googlebot and a human visitor. In 2026, a page also gets read – and cited, or not – by AI Overviews, ChatGPT, Perplexity, and Gemini. That’s a new audience with its own checklist.
I added a section for it:
- AI Overview / SGE visibility — structuring key sections as direct, quotable answers near the top of the page, so an AI Overview can lift them cleanly.
- Answer Engine Optimization (AEO) — writing for citation by ChatGPT, Perplexity and Copilot, not just ranking on a SERP.
- Entity and brand consistency — keeping brand facts (name, address, claims) consistent across the web, because that’s the grounding data these models pull from.
- AI crawler rules in robots.txt — explicit allow/block decisions for GPTBot, Google-Extended, PerplexityBot and similar, which didn’t exist as a concept ten years ago.
- Author E-E-A-T signals — real bylines, bios and credentials, which Google has leaned on hard since the Helpful Content updates.
None of this replaces the technical basics. It sits on top of them. A page with broken canonicals and no schema isn’t getting cited by an AI Overview no matter how well the paragraph is written.
This matters in practice, not just on paper. I run this checklist across every audit for my 25-person SEO team, covering campaigns in the US, UK, Canada and Australia – if one auditor is still checking keyword density while another is checking Core Web Vitals, client reports stop being consistent. A checklist only earns its place when everyone on the team is checking the same things, updated the same way.
What I would do with a checklist like this
- Date-check it once a year, minimum. If a line references a specific tool or metric by name, that name is the first thing to expire.
- Tag every item by why it matters, not just what to check. “Comment” columns that just hold a tool link don’t help a junior auditor understand the reasoning – they just copy the check without knowing when it stops applying.
- Separate “always check” from “situational.” Not every client needs hreflang or JS-rendering checks. Bloating the core checklist with edge cases makes people skim past the items that actually matter every time.
- Build in a status column — still valid, updated, new, remove – so the next time you revisit it, you’re not starting from a blank read-through of 50 rows.
What I would not do
I would not throw out an old checklist just because it’s old. The mistake isn’t that I built this in 2015 – it’s that I hadn’t looked at it critically since. Most of the technical SEO fundamentals in it are still exactly right. The fix was updating the 20% that had quietly gone stale, not rewriting the 80% that hadn’t.
Using AI to run these checks? Read why I never trust an AI audit without verifying it: Can AI Replace Manual SEO? 15 Years of Audits.
FAQ
Yes – arguably more, not less. AI Overviews and answer engines still pull from indexed, crawlable, well-structured pages. Weak on-page SEO means you don’t get crawled well enough to be cited in the first place.
Fixed keyword density and heading repeat-count rules. Both push toward mechanical, stuffed writing that current algorithms are specifically built to demote.
Not always. Some, like the mirror-sites check, are better merged into a broader item (duplicate content across domains) than deleted outright – the underlying risk hasn’t disappeared, it just shows up differently now.
At least once a year. Anything that names a specific tool, metric, or platform (Search Console, GA4, Core Web Vitals) is the first thing to go stale when that tool changes.
Get This Checklist as a Free PDF
Chandan Gupta is an SEO expert with 15+ years of experience helping businesses improve search visibility. He specializes in technical SEO, local SEO, and on-page optimization. Read his full story →
