Quick answer: A brand that ranks on Google but never gets named by ChatGPT, Copilot, Perplexity, Gemini or AI Overviews usually has one of five problems: the engines cannot identify the business as a clear entity, its pages answer nothing directly, no third-party source corroborates it, it is thin or blocked in Bing’s index, or its authority is split across weak duplicate pages. All five are fixable; most are diagnosable in half an hour.
On this page
- How do AI assistants decide which brands to name?
- Cause 1: The engine cannot tell who you are
- Cause 2: Your pages do not answer anything directly
- Cause 3: Nobody else says you exist
- Cause 4: You are invisible in Bing or blocking the crawlers
- Cause 5: Five weak pages instead of one strong one
- How do I diagnose it in 30 minutes?
- Symptom, cause, fix and who does it
- What does an AEO/GEO programme do about each cause?
- Frequently asked questions
How do AI assistants decide which brands to name?
An assistant names a brand when three things line up: it can retrieve a page about you, that page helps it answer the question, and it is confident enough about who you are to repeat the name. Miss any one and you are simply absent, with no error message and no ranking drop to warn you.
The plumbing differs by engine. ChatGPT search and Microsoft Copilot ground their answers largely on Bing’s index. Google AI Overviews, AI Mode and Gemini draw on Google’s index. Perplexity runs its own crawler on top of third-party indexes. So “we rank on Google” and “ChatGPT never mentions us” can both be true at once.
The process is otherwise consistent: the assistant rewrites the question into narrower searches (Bing calls these grounding queries), fetches matching pages, extracts passages that answer each sub-question, then decides which entities to name.
Cause 1: The engine cannot tell who you are
If your company name, address, category and description differ between your website, LinkedIn, Google Business Profile and the directories, the engine treats you as a low-confidence entity and leaves you out rather than risk naming the wrong business. This is the most common cause we see in Singapore, especially for firms that have rebranded, moved office or trade under a name different from their ACRA-registered one.
Two fixes do most of the work. Pick one canonical name, address format and category, and make every public profile match word for word. Then add Organization schema to the site (name, URL, logo, address, contact, and sameAs links to your LinkedIn, Google and directory profiles). That schema is the machine-readable sentence “these profiles are all the same company”.
Cause 2: Your pages do not answer anything directly
Assistants extract passages, not pages. A service page that opens with three paragraphs about your philosophy gives the engine nothing quotable, so it quotes a competitor who wrote the answer in the first sentence. This is behind most “we have great content but never get cited” complaints.
Before: “At Acme, we believe every business deserves a digital partner who truly understands its journey…” After: “Acme is a Singapore accounting firm that files corporate tax and GST returns for small companies, on fixed monthly fees. Most clients onboard in two weeks.” The second version states what you are, who you serve, how you charge and how long it takes, in two sentences the engine can lift straight into an answer.
Structurally that means answer-first leads, H2s phrased as the questions buyers actually ask, and a visible FAQ mirrored in FAQPage schema. Our explainer on what answer engine optimisation is covers the craft in detail.
Cause 3: Nobody else says you exist
Engines are reluctant to recommend a business that only its own website vouches for. Listicles, reviews, directories and press turn “a company that claims to do X” into “a company known for X”. If every competitor in the answer appears in two or three “best X in Singapore” round-ups and you appear in none, the gap is here.
What counts: independent comparison articles in your category, a steady flow of specific Google reviews, consistent listings on reputable directories, and genuine press or industry-body mentions. What does not: paid links from irrelevant sites, bulk submissions to low-quality aggregators, and review bursts that look manufactured. This is the slowest cause to fix because it depends on other people, which is why it is worth starting early.
Cause 4: You are invisible in Bing or blocking the crawlers
Because ChatGPT and Copilot ground on Bing, a site that is thin or unverified in Bing is thin in ChatGPT no matter how well it does in Google. And a single robots.txt line can block the AI crawlers entirely. This is the quickest cause to confirm and the one most often missed, because nobody in marketing looks at Bing.
Check three things: is the site verified in Bing Webmaster Tools with the sitemap submitted and pages indexed; are you using IndexNow so updates reach Bing in hours rather than weeks; and does robots.txt (or your CDN’s bot rules) block the agents that power AI answers? The crawlers to look for:
- OAI-SearchBot — surfaces and cites pages in ChatGPT search. Blocking it removes you from ChatGPT answers.
- ChatGPT-User — fetches a page live when a user’s prompt requires it.
- GPTBot — OpenAI’s training crawler. Many sites block it deliberately; that is a separate decision from search visibility.
- ClaudeBot and PerplexityBot — Anthropic’s and Perplexity’s crawlers.
- Google-Extended — controls Gemini’s use of your content. AI Overviews come from the normal Googlebot crawl, so this token neither adds nor removes you there.
- Bingbot — the ordinary Bing crawler that Copilot and ChatGPT ultimately rely on.
We have seen this lever move fastest. When we published eight guide pages on 8 July 2026 and pushed them through IndexNow, our Bing Copilot citations went from roughly 34 a day to roughly 2,066 a day within 24 hours. Those are citations, not traffic or leads, and our category is unusually AI-search-heavy, but the direction holds: if Bing cannot see you, nothing downstream can. Our guide to tracking AI citations explains how to read the report.
Cause 5: Five weak pages instead of one strong one
When several overlapping pages cover the same topic, none is the obvious best source, so the engine cites none of them or picks a competitor’s single definitive page. AI retrieval punishes cannibalisation harder than classic search did.
The pattern is familiar: a service page, a pricing page, two blog posts and a landing page that each half-explain the same offer, with no schema and internal links pointing in circles. Consolidate to one canonical page per question, redirect the rest, and make that page complete: what, for whom, how, how much, how long, and the obvious objections.
How do I diagnose it in 30 minutes?
No paid tools are needed for a first pass. Work through these in order and note where the first failure appears.
- Ask the five assistants (10 min). Put the same prompts into ChatGPT, Copilot, Perplexity, Gemini and Google: “best [category] in Singapore”, “recommend a [category] for [specific need]”, and “is [your brand] a good [category]?”. Record who is named and which URLs are cited. Confusion on the third prompt confirms Cause 1.
- Bing Webmaster Tools (5 min). Verify the site if needed and open the AI Performance report. Near-zero citations alongside a healthy Google Search Console points to Cause 4.
- Google Search Console (5 min). Filter Performance to your key topic. Three or more URLs sharing impressions for one question is Cause 5.
- robots.txt (3 min). Search yourdomain/robots.txt for the crawler names above, and check Cloudflare or WAF bot rules, which block silently.
- Schema validator (5 min). Run your homepage and main service page through Google’s Rich Results Test or the Schema.org validator. Missing Organization schema, or FAQ schema that does not match the visible text, is Cause 1 or 2.
- Read your top page as a stranger (2 min). If you cannot say what the company does, for whom and at what price by the end of the first paragraph, that is Cause 2.
Symptom, cause, fix and who does it
The same symptom can have several causes, so treat this as a shortlist rather than a verdict.
| Symptom | Likely cause | Fix | Who does it |
|---|---|---|---|
| Assistant describes you as a different company, or hedges about what you do | 1 — entity confusion | One canonical name/address/category everywhere; Organization schema with sameAs | Marketing + developer, one-off project |
| Assistant cites a competitor page that says what yours says, only faster | 2 — no direct answers | Answer-first leads, question H2s, FAQ mirrored in schema | Content owner; agency if volume is the bottleneck |
| Competitors appear in every round-up; you appear in none | 3 — no corroboration | Earn listicle and directory mentions; publish cases; build reviews | PR / partnerships; slowest lever |
| Google healthy, Copilot and ChatGPT empty | 4 — Bing blind spot | Verify in Bing Webmaster, submit sitemap, enable IndexNow | Developer; hours, not weeks |
| Perplexity and ChatGPT never cite you despite a large site | 4 — robots.txt or WAF block | Allow OAI-SearchBot, PerplexityBot, ClaudeBot; review CDN bot rules | Developer / IT |
| Cited occasionally, but always a different weak page | 5 — cannibalisation | Consolidate to one complete page per question; redirect the rest | SEO lead + content |
| Named in one engine, absent in all others | Mixed — index-specific gap | Diagnose per engine; fix the index you are missing from first | Whoever owns measurement |
Takeaway: Causes 4 and 1 are the fastest to fix and the most commonly skipped, so start there before commissioning new content.
What does an AEO/GEO programme do about each cause?
A structured programme works the five causes in sequence, because content published before the entity and index layer is fixed tends to go unseen. The order is entity and schema, then Bing and crawler access, then consolidation, then answer-first content, with corroboration running throughout.
Answer engine optimisation is the on-page work for Causes 1, 2 and 5: entity data, schema, question-led structure and consolidation. Generative engine optimisation extends that across the engines, covering Cause 4 and much of Cause 3: index health, crawler policy, citation tracking and third-party presence.
Whichever route you take, insist on a measurement loop: the Bing AI Performance report plus a monthly run of the five-assistant prompts shows whether each fix worked. For a baseline first, our free AI visibility audit runs this diagnosis on your site and tells you which cause is binding.
Frequently Asked Questions
We are in a regulated industry (finance, healthcare, education). Can AI assistants still recommend us?
Yes. Assistants routinely name licensed clinics, MAS-regulated advisers and registered schools; what they avoid is repeating claims a business cannot substantiate. Keep content factual, state your licence or registration plainly, and route outcome claims through your existing compliance process. The entity and indexing fixes in this article carry no regulatory risk.
How long does it take to start appearing once we fix these?
It depends on the binding cause. Unblocking crawlers and pushing pages through IndexNow can show in Bing’s AI Performance report within days. Entity and schema corrections usually need two to six weeks. Consolidation and new answer-first content read clearly at 30 to 60 days. Third-party corroboration is measured in months. Avoid changing everything again inside that window.
Will running Google or Meta ads make ChatGPT recommend us?
Not directly. Assistants ground their answers on organic content and indexes, not ad auctions, so paid media does not feed the retrieval step. Ads help indirectly by driving branded searches, reviews and mentions that strengthen your entity over time, but a large ad budget with an unfixed robots.txt still leaves you absent. Treat ads and AI visibility as separate budgets.
Is this different for B2B versus B2C brands?
The five causes are the same; the weighting shifts. B2C brands are more often held back by Cause 3, because consumer questions are answered from review and listicle ecosystems. B2B brands are more often held back by Causes 1 and 2, because their sites describe capabilities abstractly and their entity data is scattered across LinkedIn, partner directories and legacy domains.
Can our in-house marketing team do this without an agency?
Much of it, yes. The 30-minute diagnosis, the crawler and Bing checks, and the entity clean-up need no special tooling and are within a capable in-house team’s reach with developer support. Teams more often need help with the consolidation plan, the schema build, and sustaining answer-first content while running the measurement loop.
Are there Singapore-specific factors that affect AI visibility?
A few. Many Singapore firms trade under a brand name that differs from the ACRA entity, which multiplies entity confusion if both appear publicly. The local directory and listicle ecosystem is small, so a handful of mentions carries more weight than in a larger market. And because queries often include “Singapore” explicitly, pages that state location, service area and currency plainly tend to be retrieved ahead of regional pages.
Related reading
- AEO agency Singapore
- GEO agency Singapore
- Free AI visibility audit
- How to rank on AI search engines in Singapore
- How to rank on ChatGPT in Singapore
- How to rank in Google AI Overviews from Singapore
- How to track AI citations in Bing Webmaster Tools
- What is answer engine optimisation?
Want to know which of the five causes is yours?
AI Studio builds AI search visibility for Singapore brands — entity, schema, index health and the content that assistants actually cite. Start with a free AI Visibility Audit of your own site.