What you’ll learn
- What are AI visibility myths?
- Myth 1: Good reviews get you cited more
- Myth 2: You need listicles and how-to posts to get cited
- Myth 3: Your own pages control buying questions
- Myth 4: Reddit is the best third-party source
- Myth 5: Third parties win, so your website doesn't matter
A renovation contractor in Bukit Batok spent three months chasing five-star Google reviews after an agency promised they would get him recommended in ChatGPT. When he finally checked, ChatGPT named a competitor with a patchier review history and a long trail of forum threads about its jobs. The advice he paid for sounded logical, and the data behind AI answers says it was wrong.
Most advice about showing up in AI answers in 2026 is opinion dressed up as strategy. Over the past year, several large studies have tested those opinions against real AI answers, and a lot of them fail. This guide goes through the nine we hear most from Singapore business owners, what the data says instead, and what we'd do with a small budget.
What are AI visibility myths?
AI visibility myths are popular beliefs about how to get your brand named or cited in ChatGPT, Gemini, Perplexity or Google's AI Overviews that large studies don't support. Common ones: good reviews earn citations, listicles win, your own pages control buying questions, and an llms.txt file gets you in.
Sources: Neil Patel, AI visibility at scale, Search Engine Land on the SparkToro and Gumshoe study, Ahrefs brand visibility factors, Ahrefs llms.txt study.
One caveat before the myths. The biggest dataset here comes from enterprise brands across nine AI platforms, not Singapore SMEs. The direction of each finding should carry over to smaller firms, but treat the exact percentages as a guide rather than a forecast for your business.
Myth 1: Good reviews get you cited more
This is the one that costs SMEs the most time. Neil Patel's team looked at 9 million AI answers across nine platforms and found that sentiment correlated flat or negatively with how often a brand was cited on ChatGPT, AI Mode, Gemini, Copilot and Grok. Brands with rockier reputations were often cited more.
The likely reason is that these models retrieve on how much a brand is discussed, not on how people feel about it. A controversial brand generates threads, comparisons and news coverage. A quiet, well-liked brand generates very little text for a model to pull from.
Don't stop collecting reviews. They still drive the Google Maps pack, and they still decide whether the person who finds you calls you. Just stop expecting them to move your AI numbers on their own. Our local SEO tips for small businesses cover where reviews still earn their keep.
Need help with marketing? DMA builds and runs campaigns that grow Singapore businesses.
Free strategy call ›Myth 2: You need listicles and how-to posts to get cited
Plenty of "AI-optimised" content briefs now say every page should be a top 10 list or a step-by-step guide. In the same dataset, 64% of AI citations came from ordinary pages, not listicles, comparisons or how-tos.
Models pull the passage that best answers the prompt, whatever the page format, so a thorough service page that answers the questions buyers actually ask, with prices, timelines and specifics, often answers more prompts than a thin list written to be cited. Put the effort into making your existing pages complete before you start a new content series.
Myth 3: Your own pages control buying questions
You'd expect the company selling a product to be the best source on it, but AI models treat the seller as the least neutral voice in the room. On commercial-intent prompts, the ones people type when they're close to buying, 82% of citations went to third parties and just 3% to the brand's own pages. On informational and navigational prompts, owned pages took up to 13%.
For a Singapore business, the third parties that matter are the "best X in Singapore" roundups, local directories, industry association listings and news coverage that buyers already read. Getting into those is digital PR, and it's the work that moves bottom-of-funnel AI answers. When someone asks ChatGPT for the best aircon servicing company in Tampines, it is reading those roundups, not your pricing page.
Before you pay for any AI visibility work, search Google for "best [your service] Singapore" and list the ten articles on page one. If you're missing from most of them, that gap explains more of your AI visibility than anything on your own site.
Myth 4: Reddit is the best third-party source
Reddit gets most of the attention in AI search circles. In the Neil Patel data, YouTube delivered the biggest visibility lift at 2.8 times, ahead of LinkedIn at 2.7 times, Reddit at 2.4 times, G2 at 1.9 times and Gartner at 1.6 times.
A video transcript is a long, well-structured, spoken document, which is exactly what these systems retrieve well. Most SMEs treat YouTube as a brand channel and never think about the words in the video. If you already film project walkthroughs or explainer clips, make sure each one has a full, accurate transcript, a title that names the service and the location, and your brand said out loud more than once.
Myth 5: Third parties win, so your website doesn't matter
Some brands read the third-party numbers and cut back on their own site, which gets the data backwards. Owned citations are rarer, but when a brand's own content did get cited, its AI visibility rate rose about five times, the strongest lift anywhere in the dataset. Owned pages also stayed cited three to nine times longer than third-party citations.
"Fundamentals beat hacks, and data beats opinion."
Neil Patel, AI visibility at scale
So the site still matters, as long as it's the kind of source a model can quote, with specific and current answers to the questions your buyers ask. If you want the full method, our guide to generative engine optimisation walks through it engine by engine.
Myth 6: Once you're cited, you stay cited
Agencies love a screenshot of a client named in ChatGPT. The data says to be careful with it. Up to 58% of citations in the dataset never reappeared. More than half of what a brand won, it won once.
Judge your content on how long it keeps getting cited rather than on one good week, which means checking the same prompts every week for the first month and monthly after that. When a page drops out, refresh it with newer figures before you write something new. Original numbers help most here, which is why we wrote up how original research earns AI citations.
Myth 7: Every citation builds your brand
A citation means the model used your page. It doesn't mean the answer said your name. Up to 75% of AI answers that cited a brand's page never named the brand. Your research did the work, and someone else's name sat in the answer.
That changes what you report. A rising citation count on its own is a vanity number. Pair it with share of voice (how often you're named across a fixed set of prompts), branded search in Search Console and direct traffic. Our guide to AI search attribution shows how to set those up in GA4 and Search Console.
Myth 8: You can track an AI ranking like a Google ranking
Several tools now sell an "AI rank" for your brand. SparkToro and Gumshoe had 600 volunteers run 12 recommendation prompts through ChatGPT, Claude and Google's AI tools 2,961 times. There was less than a 1 in 100 chance that ChatGPT or Google's AI would return the same list of brands twice, and the order and length of the list changed too.
So there is no stable position to track. What is stable is how often you appear across many runs. And when that share swings, look at your own brand before you blame the platform. In the Neil Patel data, volatility between brands ranged from 0.04 to 1.0, against 0.08 to 0.32 between platforms. On the same platform in the same week, some brands held steady and others swung. The steady ones had deep topical coverage and strong third-party mentions.
Myth 9: Each AI platform needs its own playbook, and quick fixes work
Pitches often promise one tactic for ChatGPT, another for Gemini and a schema trick for AI Overviews, yet brands that did well on one platform did well on all of them, and the top performers were about four times steadier than the rest everywhere at once.
The quick fixes don't hold up either. Ahrefs checked 137,210 domains in May 2026 and found 97% of llms.txt files got no requests at all. In a separate study of 75,000 brands, branded web mentions correlated with AI Overview visibility at 0.664, against 0.218 for backlinks. Being talked about across the web beats any file you can upload.
Myths against the data, side by side
Here is the short version, with what we'd do instead for a Singapore SME.
| Myth | What the data shows | What to do instead |
|---|---|---|
| Good reviews earn citations | Sentiment flat or negative against citations on 5 platforms | Keep reviews for conversion; build mentions for AI |
| Listicles and how-tos win | 64% of citations come from ordinary pages | Make service pages complete and specific |
| Own pages control buying prompts | 82% third party, 3% owned on commercial prompts | Get into local "best of" roundups and directories |
| Reddit is the top source | YouTube 2.8x, LinkedIn 2.7x, Reddit 2.4x | Add full transcripts to videos you already make |
| Website doesn't matter | Owned citations lift visibility about 5x and last 3 to 9x longer | Keep investing in pages buyers ask about |
| Citations stick | Up to 58% never reappear | Re-check monthly and refresh pages that drop |
| Every citation builds brand | Up to 75% of citing answers don't name the brand | Report share of voice and branded search |
| AI rankings can be tracked | Under 1% chance of the same list twice | Track how often you appear across repeated runs |
| Per-platform playbooks and llms.txt | Winners win everywhere; 97% of llms.txt files unread | Spend on fundamentals that transfer |
A 30-minute monthly check for Singapore SMEs
You don't need a paid platform to find out whether AI answers mention you. Here's a routine that works at SME scale.
- Write 15 to 20 prompts a buyer would type, in their words. For example, "which accounting firm in Singapore handles small company tax filing" or "best aircon servicing in the east".
- Run each prompt three times in ChatGPT, Gemini and Google AI Mode, in a logged-out or private window. Three runs matter because the answers change every time.
- Record two things per run: were you named, and which competitors and websites were named or cited.
- Work out your share: runs where you were named, divided by total runs. That's the number to watch month to month.
- List the third-party sites that keep appearing. Those are your PR and listing targets for the next quarter.
| Action | Time each month | Cost | Myth it answers |
|---|---|---|---|
| Run the prompt check above | 30 minutes | Free | 8, 6 |
| Pitch two local roundups or directories | 2 hours | Free to low | 3, 9 |
| Fill gaps on one service page (prices, timelines, FAQs) | 2 to 3 hours | Writing time | 2, 5 |
| Add transcripts to existing videos | 1 hour | Free | 4 |
| Check branded search in Search Console | 15 minutes | Free | 7 |
Keep the prompt list fixed for at least three months, or the trend means nothing. Add new prompts in a separate column instead of replacing old ones.
How to read an AI visibility pitch
Now that the myths are clear, the pitches get easier to judge. Be wary of anyone who guarantees a position in ChatGPT, reports a single AI rank without repeated runs, or sells an llms.txt file or a schema package as the main fix. Ask instead how they will measure share of voice, which third-party sites they will target and how they will refresh pages that drop out. Those answers tell you whether they've read the data. If you'd like a second opinion on a proposal, our AI SEO team in Singapore will look at it.
Frequently asked questions
Do good Google reviews help my business show up in ChatGPT?
Not directly. A 9-million-answer study found sentiment had a flat or negative link with how often brands were cited on five AI platforms. Reviews still matter for Google Maps and for converting the people who find you, so keep collecting them.
Does an llms.txt file help AI tools find my site?
Almost never, based on current data. Ahrefs looked at 137,210 domains in May 2026 and found 97% of llms.txt files received no requests at all. Your time is better spent on complete service pages and mentions on other sites.
Can I track my brand's ranking in ChatGPT?
Not as a fixed position. SparkToro and Gumshoe found less than a 1 in 100 chance that ChatGPT or Google's AI gives the same brand list twice. Track how often you're named across repeated runs of the same prompts instead.
What matters most for AI visibility?
Being discussed on other websites. Ahrefs found branded web mentions correlated with AI Overview visibility at 0.664, three times the figure for backlinks. On buying-intent prompts, 82% of citations went to third-party sites.
Is AI visibility only for big brands?
No. Small businesses can win narrow, local, high-intent prompts where big brands have little coverage, such as a service in a specific Singapore neighbourhood, and the monthly routine in this guide takes about four hours whatever your size.
Get an honest read on your AI visibility
We run the prompt check, map the third-party sites your buyers' AI answers rely on, and fix the pages that should be getting cited. If you'd like to know where you stand before spending on it, talk to our team.
