Key Takeaways
When ChatGPT, Gemini, or Perplexity are asked for recommendations, Philippine brands get named in 83% of the answers they should be part of; US brands managed to only be in 40%. The gap means answer slots in the Philippines are still up for grabs, but a strong website does not guarantee a mention.
- Philippine brands are mentioned in 83% of the AI answers they should be part of, while US brands are in 40%.
- Gemini mentions brands most often (63%). ChatGPT mentions them least (45%).
- Only 5% of the sources AI links to when describing a Philippine category belong to the brand itself.
- A brand with a rock-bottom link score was answered every time, but a decades-old US site with a strong link profile was not.
Somewhere in Metro Manila right now, a business owner is typing this into ChatGPT: “best business loan provider in the Philippines for a small company.” Three seconds later, they have a shortlist of names.
Here’s the uncomfortable question: is your brand on it?
We decided to find out for real. Over three weeks, we generated 1,146 real buyer questions for 26 companies (16 Philippine brands and 10 US brands run through the same process as a comparison group), screened them through AI, then took the 255 questions each brand genuinely deserved to win and fired those at ChatGPT, Gemini, and Perplexity.
Then, we read every single answer. To our knowledge, nobody has done this for the Philippine market before.
The result? To keep it short: Philippine brands showed up in 83% of the AI answers like ChatGPT that they should be winning; US brands managed just 40%.
That gap is the whole story, and the window to claim that territory is closing. Here’s what the data shows and what to do about it.
First, What is AI Visibility?
When someone asks ChatGPT, Gemini, or Perplexity for a recommendation, they don’t get 10 links. They get an answer that names a handful of companies and cites its sources. Being one of those companies is what AI visibility is. Everyone else is invisible because this isn’t a SERP with two or more pages.
The work of getting your brand into those answers is called answer engine optimization (AEO). If classic SEO is a fight for a position on a results page, AEO is a fight to be one of the few names the AI says out loud.
In this study, we’ll use two terms throughout. A winnable question is one where the brand genuinely should have come up because it sells that thing, in that market, to that buyer. A citation is a source the AI links to when it answers, the way a footnote works.
Key Statistics
These are the numbers worth stealing for your next deck (a link back here would be nice):
- 1,146 locked buyer prompts screened, 255 winnable questions judged across three engines, and 1,911 stored AI answers, done between June 4 and August 3, 2026
- Philippine brands were mentioned in 83.2% of the AI answers they should have been part of (129 of 155 questions); US brands are at 40%.
- Philippine brands were absent from every single engine on 1 out of every 6 commercial questions they should have shown up for. For US brands, it was 3 out of every 5.
- Gemini is the most brand-friendly engine for Philippine queries (62.6% mention rate); Perplexity (57.4%) and ChatGPT (45.2%) follow.
- Of the 7,308 sources AI cited in Philippine answers, only 5% were websites the brand actually owns.
- In US answers, competitor websites take 60.7% of those sources. In Philippine answers, they only hold 36%.
- A strong website had no measurable relationship with AI visibility. One brand with a rock-bottom link score was mentioned in every answer. A decades-old US site with a strong link profile was never mentioned at all.
- We ran one brand 11 times over five weeks. Of the questions we repeated, 90% of those that came back empty stayed empty every time.
| Engine | Philippine brands | US brands |
|---|---|---|
| Gemini | 62.6% (97 of 155) | 20% (20 of 100) |
| Perplexity | 57.4% (89 of 155) | 28% (28 of 100) |
| ChatGPT | 45.2% (70 of 155) | 26% (26 of 100) |
| Any engine | 83.2% (129 of 155) | 40% (40 of 100) |
Why We Ran This Study
Every AI visibility study we could find was built on US brands, US questions, and US buyers. The largest of them, Semrush’s 2026 AI Visibility Index, ran 126 million prompts, every one of them from the US. Helpful if you’re selling software in Austin, not if you’re selling condos in Quezon City.
The measurement that does reach this market looks at traffic: how many visitors landed on your site from an AI assistant. That’s worth tracking, but it doesn’t show if the AI even says your name when your buyer asks. We call that the answer side, and nobody had measured it here.
The buyers are already there. 42.4% of Filipino internet users had used ChatGPT in the past month when Meltwater and We Are Social last measured it, against a global average of 26.5% — sixth highest of any country. What nobody had checked is whether those answers name Philippine brands.
We’ve been tracking AI visibility for our clients since 2025, and this report pools every completed study on our current process. This is what answer engine optimization looks like when you do it with receipts.
Key Findings
Finding 1: The window is open in the Philippines; the US shows it closing.
Every square is one buyer question we looked into, the kind of commercial question a brand should rightfully show up for when someone asks an AI. The dark blue squares mean no engine mentioned the brand at all.
With the same process, judging, engines, and timeline, Philippine brands did not appear on 1 out of every 6 commercial questions they should have shown up for. US brands did not appear on 3 out of every 5.
Part of that difference is how contested each market is inside AI answers—and that’s the finding. In the Philippines, most of the answer slots haven’t been claimed yet, but not in the US.
Why does that matter? Because these positions are stubborn. We ran one US brand 11 separate times over five weeks; of the questions we repeated, 90% of the ones where it was missing came back missing every single time. Being left out of an AI answer is not a bad roll of the dice that you can retry tomorrow. It holds until something about your brand or the sources talking about you changes.
Finding 2: Gemini is generous with Philippine brands, but ChatGPT is a tough crowd.
We assumed the three engines would roughly agree, but on identical Philippine questions, Gemini named the brand 63% of the time, Perplexity 57%, and ChatGPT 45%.
It’s not because ChatGPT can’t see the web. We ran it with web search turned on. ChatGPT just converts what it finds into named recommendations more cautiously than the other two. So, if you’ve only checked one engine, you haven’t measured your AI visibility accurately.
Finding 3: 95% of what AI cites about your category is not your website.
In the 7,308 citations from the judged Philippine answers we captured and classified, only 5% of the websites are owned by the brands themselves.
Read that again, because it flips the usual content playbook on its head. Publishing on your site is necessary, but the AI builds its answers mostly from pages about you: news coverage, directories, and industry roundups.
Among the US brands, competitor websites have already grabbed 61% of citations. In the Philippines, they only hold 36%. The editorial and directory layer—the part any brand can still earn its way into—carries the majority for now.
Finding 4: Domain authority does not buy AI visibility.
This one genuinely surprised us. For the longest time, we assumed in SEO that domain authority transfers: build the links, and visibility follows. Domain rating (DR) is the industry’s 0 to 100 score for how strong a website’s backlink profile is, and a higher score supposedly means a stronger site.
In AI answers, we couldn’t find that pattern at all. Philippine brands scoring under 20 were mentioned on 95% of their questions. The 20–39 group? 71%. The 40 and up group? 94%. That’s not a gradient; that’s noise.
A retailer with a rock-bottom link score owns every single answer in its category. Meanwhile, a US law firm with a decades-old domain and a strong link profile was never mentioned once.
We are not the only ones seeing this. Ahrefs ran the same question across 75,000 brands and found Domain Rating correlating between 0.27 and 0.33 with AI visibility, and raw backlink counts weaker still. What did correlate, at 0.66 to 0.71, was how often other websites mention the brand by name. Different panel, different scale, same answer.
The metrics our industry has been selling for years describe the web’s old trust system. The engines are clearly using a different one that runs on two things: whether the AI can tell exactly which company you are, and whether other websites back you up.
Finding 5: When AI fails a Philippine brand, it fails in one of three ways.
We read every failed answer, and the same three problems show up for brands that didn’t get mentioned. They hit Philippine companies especially hard, because we love conglomerate naming here.
Mode 1: AI answers about a different company with the same name
A brand can look perfectly healthy because it turns up in AI answers about its category, but every question asked about the brand by name gets replies with a different company that happens to share that name, often one based abroad. The buyer never notices; neither does the brand because the headline number looks fine.
Mode 2: AI refuses to commit to who you are
We asked ChatGPT what a specific Philippine financing company does, and it declined to answer. We ran these tests through the engines’ APIs rather than a chat window, which is why we can quote the response exactly, with the model version and timestamp attached.
This is a real, stored answer from our fieldwork:
Prompt: What does [Brand] do?
Answer:“There are several firms called ‘[Brand].’ Which one do you mean?”
On that same brand’s questions, ChatGPT named it 25% of the time, and Gemini managed 62.5%. That’s not a content problem, but rather an identity problem. The good news is that it’s fixable.
Mode 3: The subsidiary gets swallowed by the parent brand
Several property brands in our study hit this. Ask about the subsidiary, and the AI answers about the parent company or mentions the subsidiary only as an afterthought. The buyer hears the parent and remembers that. The gravity is real, and no report will ever show it to you.
None of the three failures appear in rank trackers or analytics. They only appear when you read the actual answers, which is why we keep every single one.
It’s worth checking whether one of these three is happening to you. Ask ChatGPT what your company does, and see if it can answer without hedging. If it stumbles, our free AI visibility scan will tell you which of the three issues you’re having.
What Should Philippine Brands Do About This?
Three things, in order of impact.
1. Make sure the AI knows who you are
Before you optimize, ask each engine what your brand does. If it hesitates, describes a namesake, or folds you into your parent company, fix that first. A definitive identity page, one consistent name everywhere, and structured data that spells out who you are. In our data, the one brand ChatGPT could not disambiguate was named on a quarter of its questions there, against nearly two-thirds on Gemini.
2. Chase citations, not just rankings
Remember: 95% of what AI cites is not the brand’s own website. The editorial and directory layer that feeds Philippine answers is mostly unclaimed, so every mention you earn in a publication or listing the engines trust converts into an AI answer presence. The industry has been calling it digital PR for years, except now it has a scoreboard.
3. Answer your buyers’ questions before your competitors do
The engines assemble answers from pages where the answer already exists. Every commercial question in your category that nobody has answered well is a slot with your name in it. Publish the direct, specific answer on your site and in third-party coverage while it’s still open. The US brands show you what it looks like when you wait.
What’s Next in This Study
Here’s something that kept us up at night: AI answers aren’t deterministic. Ask the same question twice, and you can get different brands. On one brand, we ran 11 rounds over five weeks; of the questions we repeated, 90% of the unanswered ones stayed unanswered every time. Not a single win held. Absence is sticky, but presence is fragile.
In the next installment, we’ll re-run all 26 brands three times and find the flip rates for each engine. This is a living study. The panel grows, and the numbers update. If you want to see how your category is settling, keep this page bookmarked.
So, Is Your Brand in the Answer?
Every number in this study started with a simple question: when your buyer asks an AI about your category, does your name come up? You can find out on your own in a few minutes. Our free AI visibility scan runs your category’s buyer questions through the AI engines and shows you where you actually appear, and what is getting recommended instead.
If the scan turns up a problem worth fixing, that is what our answer engine optimization service does: the full process, your buyers’ real questions judged with quoted evidence, entity problems diagnosed, fix plan attached. Either way, the window in your category will not stay open forever. Better to know where you stand while it is.
Frequently Asked Questions
Not that we can find. Across our 26 brands, the correlation between Ahrefs Domain Rating and AI visibility was basically zero. One brand with a rock-bottom Domain Rating got mentioned in every single answer we judged. A US firm with a decades-old domain and a strong link profile never got mentioned once. What the engines seem to reward is a clear, unambiguous entity and other websites vouching for you, not your backlink profile.
For Philippine questions, Gemini named the most brands, at 62.6% of the questions we judged. Perplexity came in at 57.4% and ChatGPT at 45.2%. In our US group, Perplexity was 28%, ChatGPT was 26%, and Gemini was 20%. The takeaway is that each engine behaves like a different editor, so checking only one gives you a false picture.
First, make sure the AI knows you: a consistent brand name everywhere, a clear identity page, and structured data. Second, get other websites talking about you, because 95% of what AI cites is not the brand’s website. Third, publish direct answers to the commercial questions your buyers actually ask, since the engines build answers from pages where the answer already exists.
Fewer Philippine companies are competing for those answers so far. In the US, the incumbent brands and the websites that get cited have already locked in most of the answer slots. In the Philippines, most slots are still up for grabs. That will not stay true forever, which is exactly why we ran this study.
Methodology and Honesty Notes
- The panel. 16 Philippine brands across 14 industries, alongside 10 US brands across 10 industries, ran through the identical process as a comparison group. One of the 16 Philippine brands is our own site; excluding it, the Philippine any-engine rate is 82.5% rather than 83.2%.
- The instrument. 1,146 grounded buyer prompts, 645 Philippine and 501 US. We generate them per brand from its real services, locations, and audience, then lock them before anything gets fired.
- The screen. Every prompt went through ChatGPT (gpt-5-mini, web search on) to establish which questions each brand had a genuine claim to win. That produced the judged set: 155 Philippine and 100 US questions.
- The run. Those winnable questions were fired at all three engines: ChatGPT, Google Gemini (2.5 Flash), and Perplexity (Sonar). We judged every answer for brand presence and captured every citation (7,308 Philippine; 7,169 US).
- The receipts. 1,911 stored answers in total: 1,146 from screening, 765 from the judged runs. Fieldwork ran June 4 to August 3, 2026, most of it in the final three weeks. Domain ages come from registry and Internet Archive records. Authority and traffic metrics come from Ahrefs.
Now, the honest caveats:
- Our prompts are brand-specific buyer questions, not a random sample of an entire industry. So, read every rate as “on the questions a real buyer would ask about this brand’s category,” not “across the whole internet.”
- Prompt mixes vary by brand, and part of the Philippine vs. US gap reflects how contested each category is inside AI answers. We’d argue that’s the finding rather than a flaw, but know that it’s baked in.
- Each question was fired once per engine in this snapshot. The repeat runs come in the next installment. All percentages carry 95% Wilson intervals, and brands are anonymized, so none of the aggregate numbers depend on knowing who they are.
- Every claim in this post traces back to a stored answer, with its full text, engine, model version, and citations. If you want to challenge a number, we can pull the receipt.
