Acclaira
Save my seat
AEOAcclaira Insights

How long does AEO take to show results in Malaysia?

One controlled experiment exists. It found 36% of new pages cited within 24 hours, a peak at day six, and decay after. None of it appears in the ten AEO timelines the engines cite today.

Dan Duar15 August 202616 min read
A black hourglass lying on its side on a cream surface, its champagne-gold sand already run out and spilled in a loose pile outside the glass, lit by one soft warm light.
Every published AEO timeline is a number without a study behind it. There is exactly one experiment, and it says something else.

Answer Engine Optimization, or AEO, is the work of getting your business named by ChatGPT, Gemini, Perplexity and Google AI Overviews. When it works it is usually fast: the one controlled experiment on record saw 36% of brand-new pages cited by Google AI Mode within 24 hours. The harder finding from the same experiment is Semrush's own flat sentence about the rest: "Many pages never became sources." One caveat before any of the numbers: they measure whether a page got cited, not whether an enquiry arrived. Those are different steps, and only the first has ever been measured.

Key takeaways

  • In a Semrush experiment, 29 of 81 newly published pages, or 36%, were cited by Google AI Mode within 24 hours (Semrush, December 2025).
  • On Google AI Mode the same experiment found citation peaked at 48 pages by day six then fell to 21 by day 30, while ChatGPT climbed steadily from eight pages to 34 (Semrush, December 2025).
  • Google itself says AI Overviews "often don't trigger", and that "Indexing and serving aren't guaranteed" (Google Search Central).
  • Of ten pages the AI engines cited when we asked this question, not one supports its timeline with a study, a sample size or a method.
  • In Malaysia, 81% of AI users use AI to search for information, but 78% still use Google as part of the same process (Vodus, 2026).

A note for Malaysian readers: AEO here means Answer Engine Optimization, not the Authorized Economic Operator scheme run by Royal Malaysian Customs.

What is the only real experiment on how long it takes?

Semrush published 81 new FAQ-style pages and tracked them daily for 30 days across Google AI Mode and ChatGPT search. Within 24 hours, 29 pages, or 36%, were cited by Google AI Mode. ChatGPT was more than three times slower, citing eight of the 81 pages in the same window (Semrush, December 2025).

That is one vendor, 81 pages, 30 days. It is the only controlled experiment we could find, though not the only measured data: a practitioner tracking four of his own sites reported that "Across four of my own sites, the first citation on Bing's AI surfaces arrived anywhere from the same day to 64 days after tracking began" (Paul Takisaki, July 2026). Four sites is not a study, but the 64-day case is a useful corrective to anyone promising days.

Two things in it should change how you read every other timeline. The first is decay, and it is Google AI Mode's alone. In Semrush's words: "On day one, 29 pages were cited. By day six, that number reached 48 pages, then fluctuated in the days that followed before eventually going down to just 21 pages by day 30." ChatGPT did the opposite, climbing from eight pages to 34, because "ChatGPT search takes longer to cite content, but once it does, those citations remain". The second is the ceiling. Even at the best single day, only 59% of the test pages were cited on Google AI Mode and 42% on ChatGPT. Semrush's own summary of that is blunt: "Many pages never became sources."

And these are best-case numbers. Semrush states plainly why. The test domain carries an Authority Score of 84, and in its own words, "This means these pages were given the best possible conditions to succeed." A new Malaysian SME domain is not that.

What the one experiment foundGoogle AI ModeChatGPT search
Cited within 24 hours29 pages (36%)8 pages
Cited at day six / seven48 pages (59%), the peak14 pages (17%)
Cited at day 3021 pages (26%)34 pages (42%)

From Semrush's 81-page experiment, published 10 December 2025, on a domain with Authority Score 84. Across the whole 30 days the best single day reached 48 pages (59%) on Google AI Mode and 34 (42%) on ChatGPT.

Why does every other article say three to six months?

Because they are repeating each other. We checked all ten sources the AI engines cite for this question, and the sharpest evidence is this. A Reddit answer, which reads like independent community experience, gives the numbers "AI mentions show up in a week or two. consistent citations show up around 2 to 3 month after. and stable ROI typically takes closer to 6 months." Those are the same three figures as the most-cited vendor page, Goodie, whose own page cites no study and links only back to itself.

That is not a consensus. It is one unsourced number set in circulation, and everyone is quoting it.

The other eight hold up no better. Stackmatix footnotes its 60-to-90-day figure to another agency's blog. B2B Marketing.Global says 3 to 6 months on one page and 60 to 90 days on another. Two of the ten, Web SEM and IDCM, contain no timeline at all despite being cited to answer a timeline question. Of the nine we could read in full, not one supports its number with a study, a sample size or a method, none mentions Malaysia, and only one concedes you might never get cited: "No legitimate agency can guarantee specific AI citations".

What do the platforms themselves actually publish?

They publish latency only for the mechanical steps they control, and none of them will state a time to citation. OpenAI says robots.txt changes take "~24 hours" to propagate (OpenAI). Perplexity says up to 24 hours (Perplexity). Google says "Crawling can take anywhere from a few days to a few weeks" (Google).

None of that is a promise about being cited, and Google is explicit that there is no promise to give.

"AI Overviews are only shown when our systems determine that it is additive to classic Search, and as such, often don't trigger." Google Search Central, AI features and your website

Read that twice, because it is the part no vendor timeline mentions. For many queries there is no AI Overview at all, so there is no slot to win and no timeline to wait out. Google adds that meeting every requirement still changes nothing about the guarantee: "Indexing and serving aren't guaranteed."

It is also worth knowing what Google says you do not need. On the AI-specific files widely sold as AEO deliverables: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn't use them." The measured evidence agrees, and we cover it in the FAQ below.

Why "how long" may be the wrong question

Because the evidence says the variable is whether you get in, not how long you wait. SISTRIX tracked 82,619 prompts over 17 weeks across six countries and found that for Google AI Overviews, "In 53% of all prompts, not a single source changes over 17 weeks." Their conclusion is the sharpest line in the literature:

"Once you're in, you stay in. If you're not in, you don't get in either." SISTRIX, AI citation drift study, 1 May 2026

On the other engines the picture inverts. SISTRIX found ChatGPT replaces as much as 74% of its sources weekly. A separate study of 530,875 citations found that asking the same question two days running changes 69% of the sources behind a typical answer, and that 84% of cited domains were cited by exactly one of the four engines tested (GetMentions, July 2026).

So being cited in week three is not a milestone you can bank. SparkToro had 600 volunteers run 12 prompts 2,961 times and concluded there is "<1 in 100 chance that ChatGPT or Google's AI, if asked 100X, will give you the same list of brands in any two responses" (SparkToro, January 2026). Their practical rule, and ours, is to measure visibility across many prompts run repeatedly, not to screenshot one answer.

One honest caveat on all of this: the research disagrees with itself about whether small domains have a chance. Surfer analysed about 5 million citation URLs and found "Domain authority doesn't appear to have any meaningful impact on the likelihood of being cited" (Surfer, July 2026). An analysis of 1,000 AI Overviews by Digital Applied found the top 1% of domains took 47% of all citations (Digital Applied, April 2026). Both are vendor studies with different samples and methods: Surfer pooled four surfaces, Digital Applied looked only at AI Overviews. Anyone telling you this is settled has not read both.

So what decides whether you get in at all?

If entry is the variable, these are the levers the evidence actually supports. Note what is missing from the list: publishing more. Ahrefs compared 75,000 brands against millions of AI responses and found "almost no relationship between content volume— number of site pages (~0.194)—and AI visibility" (Ahrefs, December 2025).

  1. Let the crawlers in. OpenAI documents OAI-SearchBot as the crawler that decides whether a site can surface in ChatGPT's search features, and says plainly that "There is no way to guarantee top placement" (OpenAI). If it is blocked, nothing else on this list matters, and a fix propagates in about a day.
  2. Keep the server healthy. Google states: "If your server is struggling to serve bytes, our crawlers will automatically back off to avoid overloading your infrastructure, which will drop your crawl frequency" (Google).
  3. Be worth citing. Google's own guidance says creating content people find "unique, compelling, and useful will likely influence your website's presence in generative AI search in the long run more than any of the other suggestions in this guide" (Google).
  4. Exist off your own site. The most-cited domains in AI answers are not company websites. An analysis of 30 million sources found "Reddit ranks as the most-cited domain in AI-generated answers, followed by YouTube and LinkedIn" (Search Engine Land, March 2026).

Point four is the uncomfortable one for anyone selling website work, including us. If the engines reach for Reddit, YouTube and LinkedIn before they reach for your site, then a programme confined to your own pages is working on the smaller lever.

Can you even measure the result?

Only partly, and the gaps matter more than the numbers. Google Analytics added an AI Assistants channel covering "ChatGPT, Gemini, Deepseek, Copilot, or Grok" but it "excludes Google's AI Overviews and AI Mode" (Google Analytics Help). Clicks from Google's own AI surfaces are counted as Organic Search instead.

Google's documentation states the same limitation for its other tool: "Sites appearing in AI features are included in the overall search traffic in Search Console" (Google Search Central). The dedicated generative-AI report exists but reports impressions only, and Google says "Not all properties have access to the report, as we're rolling out over time" (Google Search Console Help).

Then there is the click itself. Pew Research Center watched 900 US adults across 68,879 real Google searches and found that when an AI summary appeared, users clicked a link inside it in "just 1% of all visits" (Pew Research Center, July 2025). Being named in the answer and getting a visit are different outcomes, and only one of them lands in your analytics.

Does any of this differ in Malaysia?

No published timeline data exists for Malaysia, and the audience data that does exist points somewhere useful. Ads inside AI Overviews run in Malaysia, which is Google's own confirmation the surface is live and monetised here (Google Ads Help). Malay-language AI Overviews date only from 20 May 2025 (Google, May 2025), so Malay-language coverage is about 15 months old. English coverage in Malaysia is older, and most Malaysians still prompt in English.

The local behaviour is the part worth planning around. A census-stratified study of 2,556 Malaysian adults found 81% of AI users use AI to search for information, 44% use AI embedded inside search engines rather than a separate chatbot, and, critically, "Even when using AI for searching for information, 78% still use Google as part of the process" (Vodus, 2026). Google still holds 92.48% of Malaysian search referrals (StatCounter, July 2026).

That combination is the practical Malaysian answer. AI use here is additive to Google, not a replacement for it, so an AEO programme that abandons ordinary search is optimising for the smaller half of the same journey.

What our own data can and cannot tell you

We can show you a measured AI channel and we cannot tell you how long it took. DNE Logistics, a Port Klang freight forwarder and our live reference, records about 116 AI Assistant visits a month from 70 people across ChatGPT, Copilot, Gemini, Claude and Perplexity, in the 30 days to 26 June 2026 (Acclaira, DNE case study).

Our own published case study says why we cannot give you the timeline: "This is not a before-and-after story. We have no earlier baseline, so we claim no growth percentage."

That is the most useful sentence in this article. The reason nobody can tell you how long AEO takes is not that the answer is secret. It is that almost nobody records day zero, so there is nothing to measure from. We did not either, on the one site we know best. It is the mistake we now design every engagement to avoid, and it costs nothing to avoid it: write down ten questions your customers ask, put them to ChatGPT, Gemini and Perplexity today, and save the answers.

Frequently asked questions

So what should I actually expect?

Expect a measured baseline in week one, and watch for movement in days rather than months. The only controlled experiment saw a third of new pages cited within 24 hours on Google AI Mode and eight of 81 on ChatGPT, then decay on Google AI Mode while ChatGPT kept climbing. Judge at 90 days against your own recorded starting point, not against a vendor's schedule.

Is AEO faster than SEO?

The published claim is yes; the evidence is thinner than the claim. Google's own position is that "optimizing for generative AI search is optimizing for the search experience, and thus still SEO" (Google Search Central). For a Malaysian business where 78% of AI searchers still use Google in the same session, treating them as one job is the safer reading.

Why did I get cited once and then disappear?

Because that is normal, not a fault. One large study found 69% of the sources behind a typical answer change from one day to the next, and SISTRIX measured ChatGPT replacing as much as 74% of its cited sources weekly. Those measure which sources an engine picks for a prompt, which is a different question from whether your page stays in the index - in the Semrush experiment ChatGPT's own citations accumulated rather than decayed. Google AI Overviews are the stickiest surface of all once you are in.

Should I buy an llms.txt file?

The evidence says no. Ahrefs found 97% of valid llms.txt files were never requested at all, a monitored 90-day test logged 84 requests out of 62,100 AI bot visits, and Google says you do not need such files to appear in its AI features. If it is a line item on a quote, ask what evidence the seller has.

How do I know it is not working?

Set the baseline first, then review at 90 days. But check the instrument before the conclusion: your analytics excludes Google AI Overviews, Search Console folds AI traffic into overall search, and readers click a link inside an AI summary only about 1% of the time. A flat zero can be a measurement gap rather than an absence of result.

Where to start

Start with the baseline, because it is the only part that cannot be recovered later. Ten questions, three engines, recorded today, repeated monthly. That is an hour of work and it converts every later argument about whether this is working from opinion into evidence.

If you would rather have it run and recorded properly, that is what Acclaira does, and every engagement starts with a free AI-search readiness audit. The other half of the question most people ask alongside this one is the money, and we set that out separately in what AEO should cost in Malaysia. Dan also runs a free one-hour AEO class on Zoom every Thursday at 9.30pm. Message us on WhatsApp, read how to check whether ChatGPT recommends your business, or see what a real monthly AEO report should contain. If you are still comparing providers, our guide to choosing an AEO agency covers the rest of the red flags.

Sources

Common questions

Frequently asked questions

So what should I actually expect?
Expect a measured baseline in week one, and watch for movement in days rather than months. The only controlled experiment saw a third of new pages cited within 24 hours on Google AI Mode and eight of 81 on ChatGPT, then decay on Google AI Mode while ChatGPT kept climbing. Judge at 90 days against your own recorded starting point, not against a vendor's schedule.
Is AEO faster than SEO?
The published claim is yes; the evidence is thinner than the claim. Google's own position is that "optimizing for generative AI search is optimizing for the search experience, and thus still SEO" (Google Search Central). For a Malaysian business where 78% of AI searchers still use Google in the same session, treating them as one job is the safer reading.
Why did I get cited once and then disappear?
Because that is normal, not a fault. One large study found 69% of the sources behind a typical answer change from one day to the next, and SISTRIX measured ChatGPT replacing as much as 74% of its cited sources weekly. Those measure which sources an engine picks for a prompt, which is a different question from whether your page stays in the index - in the Semrush experiment ChatGPT's own citations accumulated rather than decayed. Google AI Overviews are the stickiest surface of all once you are in.
Should I buy an llms.txt file?
The evidence says no. Ahrefs found 97% of valid llms.txt files were never requested at all, a monitored 90-day test logged 84 requests out of 62,100 AI bot visits, and Google says you do not need such files to appear in its AI features. If it is a line item on a quote, ask what evidence the seller has.
How do I know it is not working?
Set the baseline first, then review at 90 days. But check the instrument before the conclusion: your analytics excludes Google AI Overviews, Search Console folds AI traffic into overall search, and readers click a link inside an AI summary only about 1% of the time. A flat zero can be a measurement gap rather than an absence of result.

About the author

Dan Duar, founder of Acclaira

Dan Duar

Founder, Acclaira · Director, DNE Logistics

Dan founded Acclaira to help Malaysian SMEs get understood, trusted and recommended by AI search. He also runs DNE Logistics, a Port Klang freight and customs business, so he writes about digital growth from a business owner’s seat, not an agency’s.

Full bio →DNE Logistics

Be the answer.

Get recommended by AI search.

Acclaira builds your premium website free, then does the AEO that gets your business named by ChatGPT, Perplexity and Google AI Overviews. From RM 3,000 a month, month-to-month with no lock-in.

Join the free Thursday class Or book a free 15-min call →

No card · No sales call