Every January, somebody declares SEO dead. Every December, Malaysian businesses are still getting enquiries from Google. That gap is where most predictions about the future of SEO go wrong — they are written to be dramatic, not to be right.
Below are eight predictions, each with the reasoning shown, the data linked, and an honest note on what would prove it wrong. That last part matters. Most SEO forecasts cannot be proven wrong, which is exactly what lets them be recycled every January as fresh anxiety — the same pattern behind the SEO myths Malaysian businesses still fall for. If a prediction cannot be wrong, it is not a prediction.
Search demand is not going anywhere. Malaysia had 35.4 million internet users at the end of 2025, at 98% penetration, per DataReportal. Those people still need a dentist, a contractor, a supplier. What is changing is how many of them click a website to find one.
This is how we think about the future of SEO at ZenWeb, and it shapes the SEO work we do for Malaysian SMEs.
Source video: "The Future of SEO: Why AI Search Changes Everything (2026)" on YouTube
Quick Answer: When Google shows an AI summary, people click a result roughly half as often — 8% of visits versus 15% without one. The searches still happen and the buyers still exist. Expect fewer sessions for the same demand, which makes zero-click search a measurement problem first. Confidence: high.
We would bet the agency on this one, because it has already happened. Pew Research Center tracked 900 adults across nearly 69,000 Google searches.
| What the user did next | With AI summary | Without AI summary |
|---|---|---|
| Clicked a normal search result | 8% | 15% |
| Clicked a link inside the AI summary | 1% | Not applicable |
| Ended the browsing session | 26% | 16% |
Source: Pew Research Center, n=900 adults, March 2025.
Read the middle row twice. Only 1% of visits produced a click on a source cited inside the AI summary. Being cited is not a traffic strategy. It is a trust strategy, and it pays off outside your sessions report.
What would prove this wrong: Google giving AI answers a real click-out incentive, or regulators forcing prominent source links.
Not sure where your site stands right now?
A baseline today makes every prediction below easier to act on. See how our SEO service works →
Quick Answer: AI summaries do not hit all queries equally. They appear on 60% of question searches and 53% of ten-word-plus searches, but only 8% of one or two-word searches. The traffic at risk is mostly research traffic. Short commercial queries — the ones that fill your enquiry form — stay largely intact. Confidence: high.
This is the most under-reported finding in the debate, and it lowers the panic level considerably.
| Query type | Share producing an AI summary | Rate |
|---|---|---|
| Question words (who, what, why) | 60% | |
| Ten words or more | 53% | |
| Full sentence (noun plus verb) | 36% | |
| All searches (average) | 18% | |
| One or two words | 8% |
Source: Pew Research Center, 68,879 searches, March 2025.
Now map that onto how Malaysians search when they want to buy. “Aircond service PJ” is three words. “Dental implant Puchong price” is four. These queries end in a WhatsApp message, and they are the least likely to trigger an AI answer.
The queries at risk look like “what causes a slow website” — useful for awareness, rarely the last click before a sale. That is why blanket traffic-loss numbers mislead, and why service page SEO is getting more valuable, not less.
What would prove this wrong: Google extending AI answers into short local and commercial queries at scale.
Quick Answer: When an AI answer names three companies, the user searches the one they liked by name. Across ZenWeb’s Malaysian client base, branded search has grown from under a third of organic enquiries to over four in ten in two years. Confidence: high.
The mechanic is simple. The AI answer does the shortlisting. The user then types your company name into Google. That second search is where the enquiry comes from — and it looks like branded traffic appearing from nowhere.
| First-touch source | 2024 | 2025 | 2026 |
|---|---|---|---|
| Branded search | 31% | 36% | 43% |
| Non-branded search | 58% | 50% | 40% |
| AI assistant referral | 1% | 4% | 9% |
| Direct / unattributed | 10% | 10% | 8% |
Source: ZenWeb client sample, 500+ Malaysian SME accounts, 2024–2026. Licence.
Branded and non-branded have almost swapped places in two years. AI referrals, while growing fast, are still under one in ten — they influence far more enquiries than they deliver.
To be the name people type, you have to be the name the AI answer mentions. That is why getting cited in ChatGPT and AI Overviews stopped being a novelty project.
What would prove this wrong: branded search share flattening while non-branded holds.
Want to know your branded search trend?
It takes ten minutes to pull from your own data. Set up Search Console access first →
Quick Answer: Volume worked when a hundred thin pages earned a hundred small trickles of traffic. AI answers collapse those trickles into one summary that cites nobody in particular. By 2027 we expect fewer, deeper, genuinely original pages to outperform large content libraries. Confidence: high.
The volume playbook had a simple engine: every long-tail question was a page, every page caught a few clicks a month, multiply by 200 and you had a channel. That is what AI summaries replace — long question queries trigger AI answers most often, so those pages lose their clicks first. Volume was never a strategy. It was an arbitrage, and it is closing.
What survives is content an AI cannot produce by paraphrasing everyone else: your own numbers, your tested process, your client outcomes. This is why the old SEO mistakes cost more now — thin pages were once merely useless; now they train engines to summarise your topic without you.
What would prove this wrong: thin-content libraries still holding rankings through 2027.
Quick Answer: Position one still helps, because AI engines draw heavily on pages that already rank. But the job changes from “be the top link” to “be the sentence worth lifting”. Pages written as extractable answers get cited; pages that bury the answer at paragraph nine do not. Confidence: medium-high.
Most predictions get this backwards. Ranking does not stop mattering — it becomes the entry ticket rather than the prize. Quotable, in practice, means:
One awkward consequence: writing title tags that get clicks matters less on AI-heavy queries and more on the short commercial ones that still show blue links.
Quick Answer: Crawlability and speed are becoming table stakes that modern hosting handles by default. The growth area is entity work — making machines certain who you are, what you do, and where. Expect technical SEO to shrink into a checklist. Confidence: medium.
Twenty years ago technical SEO was a real moat. Today a competent WordPress build on decent hosting clears most of it on day one. The floor keeps rising, and standing on it stops being an advantage.
Entity SEO asks a blunter question: can a machine state with confidence that ZenWeb is a digital marketing agency in Malaysia, and back that from independent sources? That means consistent naming, structured data, and being described the same way wherever you appear. It is unglamorous, and it is the closest thing to a durable advantage in generative engine optimisation — an engine unsure who you are will not name you.
We mark this medium confidence deliberately — Google has surprised everyone before by making something technical matter again.
Not sure which of these to fix first?
Most sites cannot do everything at once, and the order matters more than the list. Compare our SEO prioritisation framework →
Quick Answer: Local is where AI helps real businesses. Engines cross-check a claimed location against reviews, photos, directories and behaviour, so thin “service area” pages age badly. A Malaysian business with a real shop, real staff and real reviews gains from this shift. Confidence: medium-high.
This is the prediction we are happiest about, because it rewards businesses that deserve it. Faking local presence used to be a spreadsheet exercise: spin up fifty city pages, swap the town name, wait.
Cross-referencing breaks that. When an engine compares your claimed Shah Alam presence against reviews written by people in Shah Alam, photos taken there, and directory records that agree, the shortcut stops working. Contradictions read as noise, and noisy entities do not get recommended.
What would prove this wrong: AI-generated fake local signals scaling faster than platforms detect them — a real risk, and why this is not high confidence.
Quick Answer: Sessions are a worse proxy for SEO value every quarter, because the influence now happens where analytics cannot see it. Expect reporting to move toward branded search growth, citation share and revenue from organic-influenced enquiries. Confidence: high, though the tooling lags badly.
A large share of AI-influenced demand is unmeasurable with today’s tools. Someone reads an AI answer that mentions you, types your name into their browser two days later, and lands as “direct”. Nothing records why. The honest response is not a better attribution model. It is a different scoreboard:
This is why calculating SEO ROI is getting harder and more necessary at once. Ask “how many sessions” and the answer looks bad. Ask “what did organic-influenced business earn” and it usually looks fine. A baseline through SEO forecasting plus tracking AI search traffic in GA4 tells you which one you are answering.
Quick Answer: Run the arithmetic and the doom story weakens. Even if AI summaries reach half of all searches by 2028, the blended click rate falls from roughly 14% to 11% — a real decline, but nothing like the collapse headlines promise.
Pew gives two anchors: an 8% click rate with an AI summary, 15% without, and 18% of searches producing one in March 2025. Blend those and extend the coverage forward.
| Measure | 2024 | 2025 | 2026 | 2027* | 2028* |
|---|---|---|---|---|---|
| Searches showing an AI summary | 9% | 18% | 30% | 42% | 52% |
| Blended click rate | 14.4% | 13.7% | 12.9% | 12.1% | 11.4% |
* Projection. Modelled on Pew Research Center click rates, 2025 baseline.
Even at 52% AI coverage, roughly one search in nine still ends in a click. That is a squeeze, not an extinction.
The fall is gentle because most searches never showed an AI summary, and the ones that do still click 8% of the time. Doom headlines quote the worst-case query and apply it to the whole channel.
Two caveats. This is a projection — the 2027 and 2028 coverage estimates are ours. And a 20% fall concentrated on your best pages hurts far more than the average suggests, which is what the real AI Overviews click-through data shows.
Quick Answer: People will still search before they buy, and reputation will still decide who they pick. So the job this year is narrow: split your rankings by query type, defend the short commercial terms, and record your branded search baseline before 2027.
Prediction pieces skip the constants because they do not sell anything. Search intent is unchanged — somebody whose pipes burst at 11pm behaves the same whether the answer comes from an AI or ten blue links. AI shortlists you; humans still choose you. And time compounds, so sites that started SEO for a new website two years ago are still ahead.
Which makes the to-do list short:
If you only do one, do the first. Nearly every business we audit is either panicking about exposure it does not have, or ignoring exposure it does.
Quick Answer: The future of SEO is fewer clicks for the same demand, brand recall doing the work rankings used to do, and measurement that has not caught up. A difficult few years for reporting, and a good few years for businesses that are genuinely good at what they do.
Strip out the drama and the picture is manageable. Clicks tighten. Research traffic goes first, commercial traffic holds longest. Brand becomes the channel everything else feeds. Businesses willing to judge SEO on revenue rather than sessions will keep investing while competitors panic and cut.
We will be wrong about some of this — that is why each prediction carries a falsifier. Check them against your own data. The future of Google Ads is going through a similar reckoning.
No. Search demand is growing, and people still need to find businesses. What is dying is the assumption that ranking automatically produces traffic. SEO becomes a discipline about being findable, quotable and memorable rather than about position alone.
Only if your traffic comes mostly from long question-style queries. Pew Research Center found AI summaries appear on 60% of question searches but just 8% of one or two-word searches. If you rank for short commercial terms like “aircond service PJ”, your exposure is limited.
No, and the framing is wrong. AI engines draw heavily on pages that already rank, so good SEO is the entry requirement for AI visibility rather than a competing option. Treat AEO and GEO as extensions of your SEO work, not replacements.
Original data or first-hand expertise published on your own site. It is the one thing an AI engine cannot generate by summarising everyone else, and it is what earns citations — your real pricing, your real process, your real results.
Each carries a falsifier. If branded search share flattens while non-branded holds, prediction three is wrong. If thin content libraries keep ranking through 2027, prediction four is wrong. Watch your own data quarterly rather than trusting any forecast, this one included.
Ready to find out how exposed your site actually is?
Book a free 30-minute strategy session — we’ll split your rankings by query type, check your branded search trend and your AI citation footprint, then give you a concrete 90-day plan with realistic traffic and enquiry targets.
Complete the form and our team will contact you to discuss your goals. Let’s grow your business.
Online
