Is SEO still worth it in 2026? What stays, what GEO changes and what you can cut
Yes, it is. Technical SEO, helpful content and the questions your audience asks all still matter. Google recommends these basics for AI Overviews and AI Mode as well. Add visibility in AI features to your reporting and check which visits actually turn into enquiries.
The sentence we have heard most often for months: "Our rankings are stable, but traffic is falling." That is not a contradiction and not a mistake in your work. It is a structural change, and compared with the years before 2025 the biggest this channel has ever seen.
This article sorts that change into four categories: what stays, what works harder, what is completely new, and what you should cut. At the end you get a set of KPIs you can put in front of a management team.
If you want to know what this work looks like as a service and when outside help pays off: → Read more: what a GEO agency does
Why stable rankings bring less traffic now
Answers shown directly on the results page can change clicks. Whether they explain a specific drop has to be checked against the queries and the pages affected.
For the causes below there is no reliable, generally valid percentage split. Check your own drop in Search Console, in Analytics and on the results page itself:
| What to check | How to check it |
|---|---|
| Answers and features on the results page | run the query in Google yourself |
| Algorithm updates | match the period against known updates |
| Competition and search intent | check the SERP and the queries |
| Content and technology | check the pages one by one |
Answers on the results page can change clicks. That does not let content and technology off the hook: which cause applies can only be checked for the individual page and query.
Losing traffic and losing revenue are not the same event. Alongside organic sessions, always report qualified enquiries, revenue and contribution to pipeline.
The new search funnel: how customers find you now
A simplified model of winning customers reads:
Google search → website → enquiry → customer
One possible path that includes further sources of information looks like this; not every buying decision runs through every stop:
Question → AI assistant → search engine → YouTube → Reddit → reviews → website → purchase
People may already have gathered information before they ever reach your website. Your analytics may not show those earlier contacts in full.
A general split between visibility and first contact is not reliably documented. Use the table as a measurement frame for your own analysis:
| Area to measure | Your own data source | Question |
|---|---|---|
| Google Search | Search Console | Which queries and pages bring qualified visits? |
| Other referrals | Analytics | Which sources lead to enquiries? |
| Direct visits | Analytics and CRM | Was the first contact triggered somewhere else? |
| Customer conversation | First call or contact form | Where did the person first hear about you? |
Search Console and Analytics do not record every stop before an enquiry. Add the question about first contact to your calls or to your contact form.
Which SEO basics still carry unchanged
None of the following is new, and none of it can be rationalised away. What held in 2020 holds unchanged in 2026:
- Technical SEO. Crawlability, a clean structure, loading time and mobile usability. Check the main content both as delivered and as rendered. Google can process JavaScript; other systems may behave differently.
- Content quality. Answer your audience's questions with original, helpful and checkable information.
- Understanding search intent. If you do not know which question sits behind a query, you answer the wrong one.
- Local basics. Google Business Profile, consistent data, reviews. → Read the full Google Business Profile guide
- Internal linking and site architecture. AI systems follow structures just as crawlers do.
If you are not confident about these five points, any GEO discussion is premature. → What our SEO work covers
Four SEO work areas for helpful content and AI search
The work areas below combine good editing with a clear, checkable description of your company. There is no evidence for a general ranking of their effect across all AI systems.
1. Sources and numbers instead of claims
The GEO study from Princeton University compared nine interventions in a controlled experiment with model answers. The results describe the visibility measured there, not a general ranking for websites or business outcomes. Read the study
| Intervention | Effect on visibility |
|---|---|
| Quotations and sources | improved an experimental visibility metric |
| Statistics | improved an experimental visibility metric |
| Expert citations | improved an experimental visibility metric |
| Clarity | improved an experimental visibility metric |
The figure of up to 115 % that gets quoted so often refers, in the study, to the visibility of sources that sat at position 5 in the test setup. It is not a promise of rankings, leads or revenue for any single website. Checkable sources are still good editing: they let readers verify what you claim.
2. Mentions on other websites
Recommendations can draw on information outside your own website. Keep your company data correct and earn independent, helpful mentions, without buying artificial ones. Google warns explicitly against inauthentic mentions. Google's guide
There is no reliable, cross-platform scale that weighs brand mentions, backlinks and content against each other. For Google, useful original content and a technically accessible website remain the documented basis.
Google: optimising for generative AI search
3. A visible author and a maintained date
A visible author and a maintained date help readers judge where a text comes from and how current it is. Update the date only when the substance changes.
4. Prices you can read without filling in a form
Publish prices or plausible price ranges where that fits your offer. They help people judge whether you are the right fit; a recommendation from an AI system does not follow from them. Our own prices are public for exactly that reason.
Four questions about visibility in AI answers
Judging AI answers raises extra questions about how sources are chosen and how any of it is measured. These four points help you sort them.
1. Rank 1 no longer means being cited
| What to check | What it means |
|---|---|
| Google visibility | Requirement for Google AI features: indexed and eligible for snippets |
| Citations in other AI products | no generally valid ranking rule is documented |
| Practical consequence | keep original, helpful content and technical accessibility in place |
A Google position is no guarantee of a mention in another AI product. For Google itself, the ordinary requirements of search and the ordinary SEO basics still apply.
2. Judge mentions by relevance, not by a monthly count
There is no reliable evidence for a fixed threshold of mentions. Check whether the independent information about your company is relevant, correct and helpful, instead of optimising a monthly number.
Orientation, not a threshold study.
3. Choose the content format to match the question
| Content type | Helpful for |
|---|---|
| Lists and overviews | getting oriented between several options |
| Specialist articles | explaining one question with experience and evidence |
| Product pages | the features, prices and limits of an offer |
| Category and hub pages | navigating to related topics |
| Home page and profile pages | placing the company and who is responsible for it |
| Discussion and community pages | first-hand experience and differing perspectives |
| How-to guides | the individual steps of a process |
| Comparison and alternative pages | transparent criteria for a decision |
Choose the format by what the reader needs to know. Google requires neither a particular text length nor the splitting of text into small answer blocks.
4. AI visibility cannot yet be measured cleanly
That is the honest wording, and it is not the same as "cannot be measured at all". What works today and what does not:
What you can measure: visits that reach your website from ChatGPT, Perplexity or Copilot show up in Google Analytics as referral traffic. So you can see that someone arrived through an AI answer. And you can check by hand whether and how AI systems name you for your most important questions.
What you cannot measure: how often your brand was *mentioned* in AI answers without anyone clicking. That is the larger part of the effect, and for the systems outside Google there is still no source for it.
For Google itself, the generative AI performance report has been in Search Console since 3 June 2026; the rollout was completed on 31 August 2026. It summarises performance from AI Overviews and AI Mode according to the reporting rules Google has published. Check the dimensions and metrics available directly in your property. Google: announcement and rollout
For ChatGPT, Perplexity and Claude the position is unchanged: monthly samples and specialised tools. The report measures Google, not AI visibility as a whole.
And the most important thing to keep in mind: someone who reads your name in an AI answer and calls you three weeks later appears in your statistics as a "direct visit". The revenue is there, only the cause is not written next to it. Plan for that instead of writing the channel off as ineffective.
The value of AI referrals and their conversion differ widely by product, query and measurement method. Measure referral traffic and the enquiries that follow in your own analytics, instead of transferring general conversion figures.
→ How to collect AI visibility data yourself today: read the practical guide
Four SEO measures you can cut
Four measures that are either ineffective or harmful.
1. Repeating keywords. Write for the question your audience has, not for the repetition. Google states that its systems understand meaning and synonyms; a fixed percentage penalty for keyword stuffing cannot be derived from the GEO study in general. Google's guide
2. llms.txt as a visibility measure. Google ignores llms.txt for Google Search, including its generative features. In May 2026 Ahrefs observed across 137,000 domains that 97 % of the files present were never fetched; that measures fetches, not later effect. SE Ranking reports no correlation with AI citations across 300,000 domains. → The full evidence · Ahrefs · SE Ranking
3. Content updates on principle. Update when the facts, the offer or the guidance require it. A new date on its own improves no measurable visibility.
4. Self-promoting "best providers" lists. They rarely help anyone decide. Google also warns that artificially generated mentions are not sustainable optimisation. Explain transparently who your offer suits and where its limits are.
Google's recommendation on text structure
Google does not require "chunking" for its generative search features. Structure your texts so that people can follow them. Separately from that, Google prohibits mass-produced pages whose main purpose is to manipulate rankings or AI answers. Google guide
Which SEO KPIs matter now, and which no longer do
First the four habits you should drop:
- Looking only at rankings
- Ignoring AI visibility
- Reporting traffic with no link to revenue
- Not measuring brand demand at all
Then the KPI set that closes the gap:
| Visibility | Business |
|---|---|
| Share of AI answers (share of AI voice) | Revenue from search |
| Brand search volume | Conversions |
| Qualified organic traffic | Customer acquisition cost |
| - | Contribution to pipeline |
Choose your KPIs by business model and by the decision in front of you. Rankings and reach alone explain neither demand nor revenue, so record visibility and business outcome side by side.
How to collect your share of AI answers without a tool, 20 fixed questions, monthly, logged: → read the practical guide
How to fold GEO into your ongoing work
There is no verifiable, generally valid maturity split across companies. The classification below works as an internal task list:
| Stage of work | What to check |
|---|---|
| Secure the basics | check indexing, technical accessibility and content |
| Measure your own starting point | document recurring questions and referral traffic |
| Test improvements | keep the change and the result clearly apart |
| Review regularly | repeat whenever something relevant changes |
A regular measurement log of your own helps more than a comparison with an unverifiable market figure. If you would rather hand the work over: → our GEO agency service
See for yourself, no strings attached
Free, no obligation, facts instead of promises: start the website check or book a first call. You get your findings by email within two working days. Then you know whether working together is worth it.
Frequently asked questions
Is SEO dead in 2026?
No. Technical accessibility, helpful content, search intent and correct company data all still matter. Google recommends the same basics for its generative search features. Which work pays off for you depends on what an actual audit of your site finds.
Why is traffic falling when rankings are stable?
Answers on the results page can change clicks. For your most important queries, check whether the results page itself has changed, then match that against Search Console and Analytics.
What is the zero click search impact on business?
Traffic, qualified enquiries and revenue are separate KPIs. Fewer clicks do not automatically mean less revenue: people can read your answer in the results, remember the name and get in touch later. Measure the three together for your own offer instead of transferring general conversion figures.
Are backlinks still important?
They remain part of normal SEO work. Google publishes no general ranking between backlinks, mentions and content for its AI features; concentrate on helpful content, technical accessibility and a real reputation.
Do companies have to update their content more often?
Only when the content actually changes. Update facts, offers and instructions when they are no longer correct; a date without substance is not an optimisation lever.
What are the SEO KPIs that matter now in a report to management?
Share of AI answers, brand search volume and qualified organic traffic on the visibility side; revenue from search, conversions, customer acquisition cost and contribution to pipeline on the business side.
Which SEO metrics after AI Overviews can you see in Google Search Console?
For Google, the generative AI performance report. Google announced it on 3 June 2026 and completed the rollout on 31 August 2026. It covers performance in AI Overviews and AI Mode under the reporting rules Google has published. It measures Google features, not mentions in ChatGPT, Perplexity or Claude.
SEO vs GEO which matters now?
Both, and not one after the other. Technical accessibility, helpful content and correct company data are the basis for classic search results and for generative features alike. There is no reliable general figure for how far other companies have come, so start with the basics and with a repeatable measurement log of your own.
What should companies do first?
For Google, check indexing and technical accessibility first. Googlebot drives Google Search; Google-Extended is a separate control for Gemini training and grounding and is not a Google search crawler.
How to read the evidence: the GEO study is a controlled experiment with model answers. The Google documentation describes Google products only. Carrying either over to other AI systems, to individual rankings or to business outcomes would be an unsupported conclusion.
Sources: Princeton GEO study · Google: optimising for generative AI search · Google: generative AI performance report · Google crawlers · Ahrefs llms.txt study · SE Ranking llms.txt study
