SEO in 2026: what changed and what still works

Short answer: SEO still works, but three things changed. AI summaries take clicks from informational queries, so commercial and local content is worth more. Getting indexed at all has become the first hurdle for smaller sites, ahead of ranking. And publishing volume without authority now actively hurts, because unindexed pages accumulate. Fix indexation, then depth, then authority.

For years SEO was a fairly settled game. Get the technical details right, publish useful pages, earn some links, climb the rankings. Much of that still holds. What is new is that a growing share of customers never see those rankings, because they asked an AI instead, and that a growing share of published pages never get indexed at all.

This covers what actually changed, what still works, and the order to address things in.

What changed, honestly

Three shifts matter more than the rest.

AI summaries absorb informational clicks. When someone asks a simple factual question, the answer frequently appears above the results and the visit never happens. Ranking first for “what is X” is worth measurably less than it was.

Indexation stopped being automatic. Publishing a page and submitting a sitemap no longer means Google will crawl it. For sites without much authority, a substantial share of pages can sit in Discovered, currently not indexed for months.

Volume turned into a liability. When indexation was near-automatic, publishing more was straightforwardly good. Now, publishing faster than a site earns authority produces a growing pile of uncrawled URLs, which dilutes crawl budget and signals a site that generates pages nobody needs.

Start with indexation, not rankings

This is the part most SEO advice still skips, and for smaller sites it is now the first problem.

Ranking discussions assume the page is in the index. Frequently it is not. In Search Console, under Indexing then Pages, the report separates pages that are indexed from those that are not, with reasons. Two are worth understanding:

Discovered, currently not indexed means Google knows the URL exists and chose not to fetch it. This is a crawl demand judgement, not a technical error. The page is not broken. Google has decided it is not worth the resource.

Crawled, currently not indexed means Google fetched the page and decided against including it. That is closer to a quality judgement.

The usual causes of the first are an absence of internal links pointing at the page, limited site authority, and publishing many similar pages quickly. A page reachable only from a sitemap is a weak signal. A page linked from several pages Google already crawls regularly is a strong one.

What to do: check what share of your pages are actually indexed before doing anything else. If it is well below 100%, internal linking and authority are your constraint, and publishing more will not help.

What still drives rankings

The fundamentals did not disappear. They became necessary rather than sufficient.

Technical health. Pages that load quickly, work properly on a phone, and can be crawled without obstruction. This is table stakes, and it is usually already fine on a modern site.

Depth that matches intent. Not length for its own sake, but genuinely answering what the searcher wanted, including the follow-up questions they have not asked yet. A 500 word page competing against thorough ones loses regardless of how well written it is.

Internal linking. Undervalued, free, and now doubly important because it drives crawling as well as relevance. Every page worth ranking should be linked from several others that already get crawled.

Authority from other sites. Still the hardest and slowest input, and still the one that sets the ceiling. Most sites that plateau have a content library that is fine and a link profile that is thin.

Experience and expertise signals. Named authors, genuine specifics, real examples, and content that demonstrates someone has actually done the thing. Generic advice assembled from other articles performs worse than it used to.

Write for questions where people need to act

Since AI summaries absorb informational clicks, the practical response is to shift weight rather than to stop publishing.

Queries where someone needs to know something simple are increasingly answered without a click. Queries where someone needs to do something, choose a provider, compare options, understand what something costs in their situation, still send people to websites, because a summary cannot complete the task.

That means cost guides, comparisons, local pages, and anything involving a decision hold value better than definitional content. Definitional content is still worth having, since it supports authority and gets cited by AI systems, but it should not be the bulk of the plan.

Where AI helps and where it hurts

AI genuinely speeds up research, technical checks, structural work and first drafts. Used that way it lets a small business publish quality content it could not otherwise afford.

The risk is not the tool, it is the pattern. AI makes publishing quickly and shallowly at scale trivially easy, and that is precisely what quality systems are built to filter out. Google’s stated position is that it judges the content rather than the production method, which is reasonable, but content produced in volume without judgement tends to look identical to every other site doing the same thing.

Two rules that keep this on the right side:

Publish only what someone would be glad to find. If the page exists to hold keywords, it will not be indexed anyway.

Keep a human deciding what is true. AI is confident about things that are wrong, and a factual error in a cost guide costs more credibility than the page earns.

The order to work in

For most small business sites, this sequence produces results fastest:

1. Confirm pages are indexed. Everything else is theoretical if they are not.
2. Fix internal linking. Every important page linked from several crawled pages. Cheap, fast, and it directly addresses the indexation problem.
3. Make business details consistent everywhere. Serves local SEO and AI citation together.
4. Deepen the pages targeting commercial queries. Fewer, better, aimed at decisions rather than definitions.
5. Earn third-party mentions. Slowest, hardest, sets the ceiling on all of the above.

Most sites attempt step four first, publish steadily for months, and never discover that steps one and two were the constraint.

For how this connects to being cited by AI tools rather than only ranking, see our guide to SEO, AEO and GEO explained.

Frequently asked questions

Does SEO still work in 2026?

Yes, but the return arrives differently. Most buying journeys still pass through search, and the fundamentals of fast pages, clear structure and genuinely useful content still determine whether you appear. What changed is that ranking first no longer guarantees the click, because AI summaries frequently answer the question above the results. Plan for high rankings producing fewer clicks than they did, and for some of the value showing up as brand familiarity rather than sessions.

Why are my pages not getting indexed by Google?

The most common cause for small sites is that Google discovered the URL but decided it was not worth crawling, which appears in Search Console as Discovered, currently not indexed. That is a crawl demand signal rather than a technical fault. It usually means the page has no internal links pointing at it, the site has limited authority, or a large number of similar pages were published quickly. Publishing more content while this is happening makes it worse rather than better.

How much content should I publish for SEO?

Less than most publishing schedules assume, and better. Volume only helps when the pages get crawled, indexed and linked, and a site publishing faster than it earns authority accumulates unindexed pages that dilute crawl budget. A useful test is whether your last twenty published pages are all indexed. If they are not, the constraint is authority and internal linking rather than output, and more publishing will not fix it.

Do AI Overviews reduce website traffic?

For informational queries, yes, noticeably. When the AI summary answers the question completely, a share of users never click any result, and that effect is strongest on simple factual searches. Commercial and local queries are less affected, because people still want to compare options, check reviews and contact a business. The practical response is to weight content towards questions where someone needs to act rather than merely know.

Is AI-written content bad for SEO?

Google’s stated position is that it judges quality and usefulness rather than how content was produced. In practice, the problem is not the tool but the pattern, because AI makes it easy to publish quickly and shallowly at scale, which is exactly what quality systems are designed to filter out. AI used to research faster and draft more efficiently is fine. AI used to produce thin pages in volume is the thing that gets sites into trouble.

Finding your actual constraint

The most useful hour you can spend on SEO is checking what share of your pages Google has indexed, and how many internal links point at the pages you most want to rank. Those two numbers explain most stalled SEO programmes, and neither requires a tool to check.

Nexiiom runs SEO, AEO and GEO programmes that start with exactly that diagnosis rather than with a content calendar. Get a free audit and we will show you where the constraint actually sits.