SEO in 2026: The Complete Guide for Businesses

Search engine optimisation is the work of improving a site's technical foundation, its content architecture and its external authority together, so that it earns visibility in search results. That is the one-sentence definition. It does not cover half of what a business needs to know in 2026, because search itself has changed.

This guide was not written to sell SEO. It was written so you know what the parts are, how long each takes, and which promises are not realistic — before you commit a budget.

Key Takeaways

• SEO is not one job. It is three layers — technical foundation, content architecture, authority — and the upper layers do not pay off until the one beneath them is fixed.
• AI summaries have measurably changed click behaviour: in Pew Research Center's browsing study, people clicked a traditional result 8% of the time when a summary appeared, against 15% when it did not.
• Promises like "first page in three months" ignore the state your site is in today. A realistic timeline depends on which layer the work has to start from.
• Ranking is an intermediate indicator, not an outcome. What deserves reporting is which query produced how many enquiries.

What does SEO actually cover?

Treating SEO as a single activity is the most common mistake. In practice there are three interdependent layers, and their order matters.

The technical foundation. Whether a search engine can crawl and index your site without friction, and whether a visitor can open a page quickly. Crawl budget, indexation rules, page speed, mobile usability and schema markup live here. Producing content while this layer is broken is like filling a bucket without patching the hole: if what you write is not indexed, the word count is irrelevant. How this layer is built into a site, and how to test it, is covered in what makes a website SEO-friendly.

Content architecture. Which page serves which search intent, and how those pages link to one another. The work here is not chasing individual keywords but building topic clusters: a pillar page with supporting content around it, wired together, with the on-page elements that still matter doing the rest. Done properly, a search engine finds the whole topic on your domain rather than a handful of separate pages.

Authority and evidence. Other sites choosing to cite you. What separates an earned link from a bought one is measurable: an earned link carries the citing publication's own traffic and topical proximity. A bought link carries neither, and carries risk — we set out what Google's own policy names as link spam, and what it says it actually rewards.

Which layer you start from depends on the state of the site. We have written out how the three run as one programme on our SEO services page.

Three of those layers have guides of their own: which schema types are actually worth implementing, local SEO and Google Business Profile, and — if you publish in more than one language — hreflang and the mistakes everyone makes.

How have AI summaries changed search?

This is the biggest change since 2025, and its effect has been measured.

Pew Research Center recorded the real browsing behaviour of more than 900 adults through a tracker installed in their browsers over March 2025. The dataset holds 68,879 distinct Google searches, of which 12,593 returned an AI summary. The result: when a summary appeared, users clicked a traditional result link only 8% of the time — against 15% when no summary appeared. Clicks on the sources cited inside the summary sat at 1% (Pew Research Center, 22 July 2025).

The same study found a second behavioural shift: people ended the browsing session entirely 26% of the time after a page with a summary, against 16% without one.

Two things follow. First, holding the same ranking no longer delivers the same traffic. Second, because the link between ranking and traffic has weakened, a report that shows only rankings no longer describes the state of your business.

It is not uniformly bleak. Seer Interactive's tracking, based on roughly 2 billion impressions, put organic CTR on AI Overview results at 2.4% in February 2026, recovered from 1.3% in December 2025 (Seer Interactive). Behaviour is settling — but not back to where it was.

Which keyword to start with is a separate decision. Why intent matters more than volume, shown with our own sector's Semrush data, is in keyword research.

How long does SEO take? The realistic answer

The honest answer is "it depends", and anyone who does not say so is selling something.

What decides it is which layer your site is stuck on:

• On a site with technical problems, the first gains can come relatively quickly. When pages that were already written but not indexed start being indexed, visibility rises without a single new word.
• On a site that is technically sound but thin, the work is writing, and writing takes time. A topic cluster takes months to settle, because a search engine does not treat you as the source on a topic on the strength of one page.
• In competitive markets, the authority layer decides it, and that layer moves slowest. Here the calendar is measured in years rather than quarters.

"First page in three months" is meaningless precisely because it does not distinguish between those three situations. A serious proposal measures the state of the site first, then gives a timeline. We collected the timelines Google's current documentation gives for each kind of work, and the core update schedule, in how long SEO takes.

What to measure, and what not to report

Ranking is an intermediate indicator. The measure that matters to a business is which query produced how many enquiries.

Worth measuring:

• Enquiries from organic sessions. Forms, calls, quote requests — whichever your conversion is.
• Performance by page group. Which cluster is working, which has stalled.
• Indexation health. How many pages are in the index, how many are out, and why.
• Visibility in AI answers. A separate thing, measured separately from classic rankings.

Not worth reporting: total keyword count, average position, impression growth. These go up while nothing changes.

SEO, GEO, or both?

Short answer: two outcomes of related work, and neither substitutes for the other — which is also why most of a GEO proposal is SEO.

Classic SEO earns you a ranked position. GEO — generative engine optimisation — aims to have you cited as a source inside an answer written by an AI. The Pew data above is exactly why the distinction matters: ranking first on a query that returns a summary no longer earns the clicks it used to. Being named inside that answer is a separate win.

Technically the two share a foundation — a crawlable site, unambiguous entity definitions, correct schema. Where they diverge is the shape of the content and the method of measurement. We have set out what GEO work covers on its own page.

What belongs in a serious SEO proposal

When comparing proposals, the thing to look at is not the price but what is actually committed to. If these are not in writing, nothing said out loud counts:

• A measurement of the starting state. A technical audit, a content inventory and a competitor comparison should be attached to the proposal. A timeline given without them is a guess.
• What gets delivered each month. "SEO work" is not a deliverable. How many pieces, which technical fixes, what reporting.
• Who will do it. The senior name in the meeting, or someone you meet after the handover.
• What gets reported. Not a list of rankings, but which query produced how many enquiries.
• The exit terms. Who holds the content, the data and the access when the contract ends.

There is also one thing that should not be in writing: a ranking guarantee. No agency controls search results; anyone guaranteeing a position either does not know that, or knows and says it anyway.

Three common mistakes

Starting everything at once. Run technical, content and authority work in the same quarter and none of them finishes — and you cannot tell which one worked. The order exists for a reason.

Measuring content by volume. Ten weak pieces a month perform worse than two good ones, and weak content can pull a site backwards.

Setting up measurement afterwards. If analytics and search console are not in place on day one, you will never see what came before — so you cannot prove the change either.

In-house, agency, or both?

All three are valid, for different things.

In-house: content production, business profile management, on-page fixes. Nobody knows the subject better than you; the writing itself usually comes out better inside.

Outside: technical audit, site architecture decisions, measurement setup and prioritisation. These layers need experience and developer access.

Mixed: the most efficient arrangement for most companies. Strategy and technical work outside, production inside.

Whichever model you pick, get one question answered plainly: who will actually do the work? The senior name in the sales meeting, or somebody you meet after the handover? In our SEO service we put that in writing from the start — the person running the programme is the person you spoke to first.

Where to start

In order — and if the budget is small, the order is the strategy:

1. Measure. Without a technical audit, a content inventory and a competitor comparison, every step after it is a guess.
2. Close the leak. Unindexed pages, broken redirects, duplicate content. Most sites gain here before a word of new content is written.
3. Start with one cluster. One topic done fully beats six done halfway.
4. Set up measurement on day one. Measurement added later never shows you what came before it.

If you want to see which searches your site is missing, ask us for a free SEO analysis — the output is a prioritised roadmap, not a pitch deck.

Conclusion

SEO still works in 2026, but it works differently. It has three layers and their order matters. AI summaries have measurably changed click behaviour, so any programme that reports rankings alone is not describing the state of your business. And a realistic timeline depends on which layer your site is stuck on today — it could be three months, it could be two years.

For a next step: if you do not know the technical state of your site, start there. If you do, and the content side is thin, start with a single cluster. If you are thinking of buying this in, SEO pricing covers the pricing models and where the figures online actually come from.