SEO Best Practices 2026: What Changed and What Didn’t

Every January, someone declares SEO dead and replaced by something with a new acronym. Every June, the same practitioners quietly go back to fixing crawl errors and writing better pages.

2026 has been a genuinely eventful year for search. But the noise has run well ahead of the substance, and the practical answer for most sites is less dramatic than the headlines suggest.

Here is what actually changed, what did not, and where to spend effort.

The Headline: Fundamentals Did Not Change, Weighting Did

Google’s own position is that optimising for AI features on Google is still SEO. There is no separate discipline, no special markup, and no required file.

What changed is where quality signals matter and how results get surfaced. The bar for what counts as quality content rose. The surface where visibility is won expanded beyond ten blue links. The measurement of success moved.

That is meaningful. It is not a reason to abandon anything that was working.

What Actually Changed in 2026

The March update was disruptive by any measure. Independent analysis using Sistrix data reported that a large majority of top-three results shifted, and roughly a quarter of pages ranking in the top ten fell out of the top hundred entirely, against about 15% after the previous December update.

If your traffic moved sharply this spring, that is the likeliest explanation, and it was a quality re-weighting rather than a penalty.

The Most Important Number in This Guide

Reported figures suggest around 97% of AI Overview citations come from pages already ranking in the top 20 organic results.

Treat the precise figure with appropriate caution, since it comes from industry analysis rather than Google. But the direction settles the strategic question people keep asking.

There is no shortcut to AI visibility that bypasses ranking. Traditional SEO is the prerequisite, not the alternative. Anyone selling AI search optimisation as a replacement for organic fundamentals is selling you a second thing to buy.

Information Gain Is the New Content Bar

The most useful way to understand the 2026 core updates is through information gain: how much genuinely new knowledge a page adds relative to what already ranks for that query.

The test is uncomfortable and clarifying. If your content disappeared tomorrow, would anyone lose access to information they could not find elsewhere?

For a large share of the web, the answer is no. Content produced at volume, optimised for keywords, and assembled from the same sources as the current top five does not clear that bar any more.

What clears it:

• Original research and proprietary data
• First-hand testing with documented method
• Case studies built from real client outcomes
• Analysis requiring access or expertise the reader lacks
• Expert commentary attributed to a named, credentialed person

The practical shift for most teams is treating first-party data as a publishing asset. Internal benchmarks, customer research, and anonymised account analyses are things you already own and rarely publish.

Usefully, this is the same standard AI systems apply when deciding what to cite. Content that is original, expert-attributed, and topically authoritative performs on both surfaces. You are not doing two jobs.

What Google Actually Says About AI Optimisation

Two clarifications worth knowing, because both contradict widely circulated advice.

You do not need an llms.txt file. Google’s AI optimisation guidance states these are not required and are not treated as special signals. Standard crawlability and indexability are sufficient.

GEO is not a separate practice for Google Search. Per Google’s own guide, optimising for AI features on Google is still SEO, and the same foundational practices improve both.

That said, structured data, self-contained answer blocks, and current statistics do appear to help non-Google engines like ChatGPT and Perplexity retrieve and cite content. So the work is not wasted. It is just not the special lever it is often sold as.

The Technical Baseline

None of this is new, which is exactly why it still separates sites.

• Crawlability and indexability before anything else. A page Google cannot reach cannot rank or be cited.
• Core Web Vitals at the documented thresholds. Reported figures suggest only around half of sites meet all three, so it remains a genuine differentiator in competitive categories.
• Clean canonicals and no contradictory signals between robots directives, canonicals, and noindex tags.
• Structured data on the templates where it applies, validated across templates rather than one sample URL.
• Mobile parity in content, not just layout.
• Internal linking that pushes authority toward commercially important pages.

Google has been clear that page experience acts as a tiebreaker rather than a primary factor. In categories where a dozen sites offer near-identical content, a tiebreaker decides the outcome.

E-E-A-T in Practice

E-E-A-T is not a score you can check. It is a set of signals you either send or do not.

Concretely:

• Named authors with real credentials and genuine bios
• Author pages linking to external profiles and published work
• Clear sourcing, with links to primary data rather than to other blogs
• Visible dates showing when content was genuinely reviewed
• Contact details, business address, and policies that a real company would have
• Third-party corroboration: reviews, citations, coverage, mentions

That last point does more work than most teams expect. A brand’s presence across the wider web feeds both traditional rankings and AI citation decisions. Strong brand signals also make a site more resilient to core updates.

Measurement Has to Change

This is where most 2026 SEO programmes are genuinely behind.

Reported CTR declines on queries with AI Overviews range widely, from around 30% to over 50% for top positions depending on the study and query type. That variance is itself informative: the impact depends heavily on query intent.

The consequence is that traditional metrics now understate your performance. Impressions can rise while clicks fall, and brand exposure can grow while analytics shows decline.

Track these alongside the usual set:

Also diversify. Personalisation in AI Mode means results vary by user, so a single tracked position describes less than it used to.

What to Ignore in 2026

• llms.txt files, per Google’s explicit guidance.
• Keyword density targets. They stopped mattering years ago.
• Adding FAQ schema for SERP features, since FAQ rich results ended in May 2026.
• Chasing every new acronym. AEO, GEO, and SEO overlap far more than the marketing suggests.
• Publishing volume as a strategy. The March update specifically de-weighted this.
• Predictions of search’s imminent death. Google’s stated position is that SEO remains the foundation for AI visibility, and the citation data supports it.

A Practical Priority Order

If you have limited capacity, work in this sequence.

1. Fix technical access. Crawlability, indexation, canonicals, Core Web Vitals.
2. Audit content against the information gain test. Consolidate or remove pages that add nothing.
3. Publish something only you could publish. One original data piece beats ten summaries.
4. Strengthen entity and author signals. Named experts, consistent brand information, external profiles.
5. Add measurement for AI surfaces. Search Console AI reports plus a manual prompt set.
6. Build third-party presence. Reviews, coverage, community, and citations.

Steps one and two usually produce the largest movement and are the least fashionable.

Mistakes That Waste 2026 Budgets

• Treating AI search as a separate channel needing a separate budget and separate content.
• Publishing more to compensate for a quality re-weighting.
• Reading a click decline as a visibility decline without checking impressions.
• Buying a tool before establishing a baseline.
• Waiting for core update recovery instead of improving content, since recoveries typically materialise around subsequent updates.
• Removing content that ranks because a tool flagged it as thin.

The Honest Summary

Search got harder to measure and no harder to do well.

The sites winning in 2026 are publishing things nobody else can publish, keeping their technical house in order, and attaching real people and real brands to their content. That was true in 2020. The difference is that doing it adequately no longer clears the bar.

Pick your five most important pages and apply the information gain test honestly. If any of them would not be missed, you have found this quarter’s work.

FAQs