What Works in SEO in 2026: A Playbook

Almost every week there is a new case study: “we did X and traffic grew 300%”. Take published SEO case studies apart and only a small share of them prove cause and effect. Seasonality, new links, an algorithm update or a dozen parallel changes are easy to credit to a single tactic. This playbook keeps what is backed by controlled tests or follows directly from how crawling, indexing and ranking work.

TL;DR

• The foundation hasn’t changed. Link equity and topical authority matter more than any hack. Authority is built with clusters of pages and dense internal linking.
• Content wins on specifics. Direct answers to real questions, numbers, tables and a short answer at the top work in Google and in AI answers alike.
• Snippets are testable. Titles and OG images move CTR, and this is one of the few areas with good A/B tests.
• Your own data is the best link source in 2026. Research, calculators and tools earn links and mentions that competitors can’t easily copy.
• A lot can be skipped. llms.txt, mass disavows without a penalty, shiny review markup and copying competitors’ old tricks do nothing or hurt.

1. How to tell a working tactic from a lucky coincidence

Before you copy someone’s case study, check it against four criteria. If it fails two of them, treat it as a hypothesis, not proof.

The gold standard in SEO experiments is split testing on groups of similar pages, as run by dedicated SEO A/B testing platforms. Keep publication bias in mind too: almost nobody publishes failed experiments. In a large sample of published SEO case studies only 2–3% report negative results, while in practice failures are far more common.

2. Foundations: link equity and topical authority

Search engines estimate how far your site is from trusted sources along the link graph and how much equity reaches you along that path. That equity is distributed by topic: a site becomes strong not “in general” but in specific query clusters. The practical takeaway: a dense network of pages around a topic beats one big page.

Clusters, not standalone articles

Pick a topic and cover it from every angle: what it is, how it works, use cases, integrations, mistakes, comparisons, questions. Dozens of connected pages on one topic eventually outrank big brands on narrow queries, because the site accumulates authority in exactly that cluster.

A ladder from easy queries to valuable ones

Build a path from low-competition informational queries to commercial ones and connect it with links. For a cleaning company: “what is deep cleaning” → “how to keep a house clean” → “when to hire a professional cleaner” → “house cleaning in Boston”. Every link on the ladder leads both users and crawlers to the page that makes money.

Internal links matter more than the sitemap

Pages with impressions and clicks get crawled most often and most deeply. When the crawler visits such a page, it queues every link on it along with the anchor context. A page the crawler only knows from sitemap.xml arrives with no context and no equity, and easily gets stuck as “Discovered – currently not indexed”. Keep the sitemap as a technical safety net, but on a young site new pages need contextual links from pages that already rank.

Internal anchors: less variety

In tests, consistent and even exact-match internal anchors performed better than dozens of variations. Link to your “SEO audit” page with the words “SEO audit”, not “here”, “read more” and five synonyms. This applies to internal links only. For backlinks, over-optimised exact-match anchors are still risky.

Key categories in the footer

A list of your main categories in the footer gives them a link from every page and noticeably boosts their equity. One caveat: think about how it looks to moderation systems. In medical niches, drug names in a sitewide footer caused problems with ad platforms because the site started to look like it was selling medication.

3. The URL as a signal: slugs, republishing and recrawls

The slug decides who you compete with

The slug is one of the strongest signals of which query a page competes for. A /seo slug puts you next to the biggest players. A slug for a narrower phrasing, say /bing-search-console instead of /bing-webmaster-tools, can land you in a results page with almost no competition. A very long slug for a rare query wins first place but little traffic. Slugs like /our-team-10-years-of-experience win nothing.

Republishing a stuck page under a new URL

All signals are tied to a specific URL. If a page has been sitting in “Crawled – currently not indexed” for ages and gets no clicks, Google barely reassesses it and ignores the authority the site has gained since. Publishing the same content with small edits under a new URL, with a 301 from the old one, often helps: the new document is evaluated from scratch with the site’s current authority. This only works if the site already has authority and internal links. A site with no authority won’t be saved by a new URL.

The lighter option: a small edit and a recrawl

If a page is indexed but hasn’t changed in a long time while the site has grown, make a meaningful edit and request indexing in Search Console. It is a supporting move: real growth still comes from links and clicks.

4. Content that ranks and gets cited by AI

Answer the questions people actually ask

The most reliable source of topics is your customers’ questions. Collect them from People Also Ask, tools like AlsoAsked, support tickets and sales calls. Format answers so they are easy to extract: a table, a short paragraph, a video, an FAQ.

H2s from page-two queries

One of the fastest wins for pages that already get traffic. In Search Console, open the Performance report, select a page and find queries with an average position of 11–30 that have impressions but few clicks. Turn those queries into H2 subheadings and answer them right below. No paid tools needed.

Separate pages for questions on a young site

On a young site, give each important question its own URL: the slug targets the query and the page’s equity isn’t spread across ten topics. Two conditions: the page answers a question people really search for, and it links to your product or service. Thousands of auto-generated FAQ pages with no internal links do the opposite.

A short answer at the top

A summary in the first lines helps both users and search. In one case, a TL;DR block lifted conversions by 33%. People who need a quick answer get it right away; people who want details keep reading.

Length isn’t quality

There is no official minimum word count. A page can rank with 100 words, a table or a video if it answers the query precisely. Google’s spam policies target scaled content without value and thin affiliate pages that only repeat other people’s offers, not short pages.

Glossaries

In B2B, finance, cybersecurity and SaaS, a glossary brings in people searching for an unfamiliar term, an acronym or even a new job title. On one B2B project, nine glossaries drive about a quarter of all organic traffic, and real leads with it.

Numbers, tables and expert quotes for AEO and GEO

The same principles apply to AI answers: specific numbers, statistics, tables, expert quotes. State results as numbers, not adjectives: “response time dropped from 4 hours to 40 minutes”, not “it got faster”. Spell out important changes explicitly: a new brand name, new terms, new features. Otherwise Google, AI assistants and forums will keep repeating outdated information.

What not to overrate. llms.txt showed no lasting effect in clean tests, and Google says plainly that it doesn’t use it. Case studies claiming “ChatGPT mentions us more” are almost always based on a single prompt, with no tracking over time and no link to traffic.

AI content: a draft, not a strategy

Google doesn’t penalise text for being generated. It penalises mass production of templated pages. The working setup: give the model all the facts (specs, use cases, limitations, compatibility) and the structure you want, let it write a draft, then have a human check the facts and edit. “Here’s a product photo, write something” produces text no better than the supplier’s description.

5. The snippet: title, price, date and OG image

Google constantly reshuffles results and watches how people click. Click data from search results is used in ranking, as evidence from the US antitrust trial against Google confirmed. Bounce rate and time on site from your analytics, however, are not direct ranking factors: a blog with a 90% bounce rate can hold its positions for years. The takeaway: the snippet matters, and a title that promises more than the page delivers loses fast, because users go back to the results and click a competitor.

The brand in the title. Google already shows the site name in the snippet. A “| Brand” suffix on every title dilutes relevance and can cause cannibalisation on brand queries. Keep the brand on contact, pricing and support pages; on the rest, test titles without it.

The OG image. Google often uses the og:image as the thumbnail in search results. Make it separate from the on-page image: a contrasting background, a person, the product, the benefit in large text. This way you control not only the headline but also the “cover” of your result.

6. Commercial pages: growth in rankings and revenue

• A visible CTA. A large branded call-to-action block on product pages delivered about +80% conversions in two tests.
• 3–4 bullets under the heading. Above the fold: who it is for, use cases, variants and colours.
• A proper description. What it is, who it is for, how to use it, what it works with. In one case, replacing supplier descriptions with clear copy brought about +85% clicks and conversions. The gain doesn’t come from AI; it comes from finally explaining the product.
• Clarity beats beauty. A plain comparison table, “us vs competitors”, with specs and prices gets pulled into the snippet and shows the key argument instantly. A beautiful page that hides the answer loses to an ugly one that shows it.

7. Links and mentions

Digital PR on your own data

The most reliable way to earn links and brand mentions in 2026 is to publish what nobody else has: research on your own data. Logs, usage stats, transactions and customer surveys become industry reports, rankings, maps and indexes. Journalists and AI models readily cite fresh, unique numbers. Competitors will read it and use it, but the gain in links and brand recognition still outweighs that.

Tools and calculators

Cost, ROI and sustainability calculators, and interactive tools for a specific job, keep earning links for years. A sustainability calculator for architects reached over 100,000 users and lots of links, especially when paired with a PR campaign. The only condition: the tool must solve a real problem.

Partnerships instead of paid links

Links from real partners (distributors, adjacent contractors, integrators) last and are hard to replicate. Help partners with their sites and swap recommendations with adjacent businesses. For local businesses, think of a plumber, an electrician and an HVAC company recommending each other.

Free link analysis

Bing Webmaster Tools shows almost the entire backlink profile of a domain, and not only yours but your competitors’ as well. You don’t need a paid tool for a quick analysis.

Disavow and “toxic” links

Links from forums, comments and random sites aren’t dangerous on their own: Google neutralises link networks rather than punishing every site they point to. A disavow makes sense in two cases: you have a manual action, or you bought links at scale yourself. A case like “we disavowed exact-match anchors and got +300% clicks” almost certainly describes a site that was already filtered for buying links. Cleaning your profile by a “toxicity score” without a penalty is, at best, a waste of time.

8. Structured data: less shine, more entities

• Careful with reviews and stars. There are known cases where sites regained rankings after removing Review and AggregateOffer markup, even though the reviews were genuine. Google is tired of abuse and errs on the side of caution.
• FAQ markup no longer earns a rich result for most sites: since 2023, FAQ rich results are shown mainly for well-known government and health sites. Write FAQs for people and AI, not for stars in the results.
• Entities over snippets. Organization and Person markup with sameAs links to official profiles helps search engines and AI connect the site to the brand and the author.
• EXIF and ImageObject are a niche tool. A travel site that tied photos to landmarks with ImageObject and geodata got about +90% impressions in Google Lens. Other tests found no effect of EXIF or LocalBusiness markup on regular rankings. Do it only if visual search matters to you.

9. Scaling: programmatic, use cases, separate sites

Use-case pages. One product can be described through tens of thousands of specific scenarios: integrations, stacks, industries, jobs to be done. In B2B and infrastructure, such projects have grown to tens of thousands of pages in many languages. It works with a clear architecture and real value on every page.

Programmatic SEO. Networks of hundreds or thousands of pages with no unique value often show a “mountain” on the graph: a quick rise and a sharp drop after the next update. Before launch, define requirements for the template: what unique data each page has and why it is more useful than the page next to it.

Separate sites and keyword domains. A separate site for a narrow niche, or a domain like “free-[product]”, works if it actually delivers what the domain promises, for example a free version of the product. In B2B, a small niche site with a few strong guides can bring in large contracts. A network of empty satellites leads to penalties.

Don’t copy competitors’ old tricks. Doorway pages, exact-match domains and fake reviews may still work for old sites by inertia. A new site using the same tricks often gets filtered or hit with a manual action right away.

10. Where good evidence is still missing

There is almost no quality research on these topics, so any firm recommendation here is closer to belief than science:

• pagination: canonical to page one, or index, follow on every page;
• content pruning on very large sites (100,000+ URLs);
• removing weak products and categories to redistribute equity.

If you do any of this, roll it out in stages, on some sections first, and record the result.

11. An SEO changelog

Without a log you can’t tell what worked. Keep a simple table and compare results 2–3 months after each change.

For a clean result, change one thing at a time and keep a control group of similar pages unchanged.

The 90-day playbook

The bottom line: SEO doesn’t need a 120-item checklist. It needs a dozen levers that actually move results, and the discipline to test each of them on your own site.