| Term | What it means |
|---|---|
| SEO: Search Engine Optimization | The practice of making a website easy for Google to find, understand and rank in search results. |
| AEO: Answer Engine Optimization | The practice of making a business the source an AI assistant chooses when it answers a customer's question directly, instead of showing a list of links to click. |
| GEO: Generative Engine Optimization | Describes the same practice as AEO under a different name; the industry treats the two as interchangeable. |
For a Malaysian SME in a services business, this is not a technical detail for a marketing team to worry about. It decides whether a customer who is looking for a service like yours ever learns your business exists.
When someone asks an AI, a business may not be mentioned
When a customer asks ChatGPT, Gemini or Google's AI Overview for a recommendation, the assistant answers directly, often without showing a single link to click. If your business was not part of that answer, the customer never learns it exists, and you never know the question was asked. This works against smaller, less well-known providers by default.
| Brand tier | Finding |
|---|---|
| Bottom two tiers (specialists & regional players) | Never appeared at all in 48-52% of answers that touched their category |
| Globally known brands | Day-1 unbranded recognition: 72.9% |
| Mid-market brands | Day-1 unbranded recognition: 43.6% |
| Niche brands | Day-1 unbranded recognition: 11.4%, roughly a 30 percentage point drop at each step down in size |
Being known and being found are also two different things: the kind of question a customer actually asks when they don't yet know who to call.
| AI engine | Named correctly when asked by name | Surfaced in general discovery questions |
|---|---|---|
| ChatGPT (gpt-4o-mini) | 99.4% | 3.32% |
| Perplexity (sonar) | 94.3% | 8.29% |
Most AI citations point to pages a business does not own
AI assistants also don't build their answers mainly from a business's own website. Across 102,025 AI responses studied, best-of style listicles alone accounted for the single largest source category measured.
| Where AI answers cite from | Share of citations |
|---|---|
| Brand's own domain | 2.9% |
| Third-party pages talking about the brand | 75.2% |
| Best-of / listicle pages (subset of the row above) | 21.0%, the single largest source category |
| Earned media (reviews, articles, comparison pieces), live AI search | 69-92% |
| Same queries typed into Google instead | 30-55% |
Search is still the foundation: AI hasn't replaced it
None of this replaces classic search engine work. It sits on top of it.
| Predictor of discovery | Perplexity | ChatGPT |
|---|---|---|
| Number of other sites linking to the company (conventional SEO signal) | r=+0.319, statistically significant | No significant predictor (answers from training data, not live search) |
| AI-specific content tactics (composite score) | No significant relationship | No significant relationship |
Search behaviour is still shifting, though. Ranking well is still necessary, but it is no longer enough on its own to guarantee a visit.
| Scenario | Effect |
|---|---|
| Google shows an AI summary above the results | Users click through to a traditional website in only 8% of visits |
| No AI summary appears | Click-through rises to 15% of visits |
| An AI Overview appears on a page where a business already ranks first | That top position's click-through rate falls by 58% |
What actually changes whether AI cites a page
Two things measurably change whether an AI assistant cites a page once it has found it.
| Factor | Finding | Source |
|---|---|---|
| Structure and section length | Rewriting pages with clearer structure (a direct answer, defined sections, one idea per block) raised citation rates from 45.0% to 52.8% in a controlled test across six AI engines | arXiv 2603.29979, 2026 |
| Local and entity signals | The primary category set on a Google Business Profile is now the single strongest local ranking factor measured | Whitespark Local Search Ranking Factors, 6 November 2025 |
| Local and entity signals | 42% of local searches end with a click inside the Google Maps results | Backlinko, updated 14 April 2025 |
This is already happening in Malaysia
This is not a US-only shift.
| Malaysian adults who say AI tools have saved them time at work | 82% | Ipsos AI Monitor 2026, Malaysia results, fieldwork 20 March–3 April 2026 |
| Malaysian adults who expect AI to change how they do their job within five years | 84% | Ipsos AI Monitor 2026, Malaysia results, fieldwork 20 March–3 April 2026 |
Malaysian customers researching a service provider are already asking an AI assistant before, or instead of, running a Google search.
What this work is worth
The return on this work is real, but smaller than most case studies suggest. The clearest evidence comes from a natural experiment that tracked AI-optimised pages against untreated pages on the same domain over 47 weeks.
| Metric | Value |
|---|---|
| Raw growth in ChatGPT referrals to the optimised pages | 6.1x |
| Growth in untreated control pages on the same site, same period, no changes made | 3.5x |
| Effect actually caused by the optimisation work, background growth subtracted | 1.82x [95% CI 1.31-2.54] |
1.82x is the number to plan around. It is a real gain, and it is roughly a third of the raw multiple that AEO case studies usually quote.
What we do about it
OtomoAI applies the same evidence to its own site: page structure, content freshness, local and entity signals, and presence on third-party pages, each checked against the findings above rather than against guesswork. The aim is narrow. When a customer in Malaysia asks an AI assistant a question your business could answer, your business should be one of the answers it gives.
Our business website design page lists what a client site includes and what it costs. The free website readiness check scores how readable any site already is to AI assistants and search crawlers.
Get in touch to talk about what this looks like for your business.
Sources
- Pew Research Center, "Google users are less likely to click on links when an AI summary appears in the results," 22 July 2025. https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/
- Ahrefs, "AI Overviews Reduce Clicks," 4 February 2026. https://ahrefs.com/blog/ai-overviews-reduce-clicks-update/
- Backlinko, "Google User Behavior" (local Maps-pack click share), updated 14 April 2025. https://backlinko.com/google-user-behavior
- Whitespark, "Local Search Ranking Factors," 6 November 2025. https://whitespark.ca/local-search-ranking-factors/
- Ipsos, "AI Monitor 2026," Malaysia results, fieldwork 20 March–3 April 2026. https://www.ipsos.com/sites/default/files/ct/news/documents/2026-06/AI%20Monitor%202026%20(English).pdf
- arXiv 2605.27439, "Prominence-Stratified Failure Modes," 2026. https://arxiv.org/abs/2605.27439
- arXiv 2606.20065, "GEO at Scale (Ranqo)," fieldwork March–May 2026. https://arxiv.org/abs/2606.20065
- arXiv 2601.00912, "The Discovery Gap: How Product Hunt Startups Vanish in LLM Organic Discovery Queries," submitted 1 January 2026. https://arxiv.org/abs/2601.00912
- arXiv 2509.08919, "GEO: How to Dominate AI Search," 2025. https://arxiv.org/abs/2509.08919
- arXiv 2603.29979, "GEO-SFE (structural feature engineering)," 2026. https://arxiv.org/abs/2603.29979
- arXiv 2606.04362, "Disentangling AEO from Platform Growth," study period July 2025–May 2026. https://arxiv.org/abs/2606.04362
·