AEO

Why AI Search Keeps Citing Your Competitors Instead of You: 7 Causes From Our Citation Research

For marketing leads who checked ChatGPT or Google AI Overviews and found competitors named while their own site ranks on Google: seven causes from our citation research, a 30 minute diagnosis, and a 30 day plan.

Why AI Search Keeps Citing Your Competitors Instead of You: 7 Causes From Our Citation Research
Share

Key Takeaways

  • A page for the exact question: 89.6% of the pages AI engines cited in our 500 query study won a citation for exactly one query.
  • Other websites name them: a vendor mentioned on 141 pages across 49 other sites earned 74.4% of its AI mentions without its own site being cited.
  • Each engine is its own channel: 94.2% of cited URLs appeared in exactly one engine, and none appeared in all three.
  • Fresher content wins: 95% of ChatGPT citations come from content published in the last 10 months (AirOps, 2025).
  • Machine readable identity: 80.8% of the pages ChatGPT cited carried schema markup, the code that tells machines who published a page, led by Organization, BreadcrumbList and Article.
  • The answer lives in the body: cited pages share only a quarter of the question's words in their headline, while their body covers 80%.
  • Rank and citation are different contests: only 15% of pages both ranked in Google's top 10 and earned an AI citation.

You asked ChatGPT which firms to shortlist in your category, and it named three competitors. Google AI Overviews did the same. Your site sits on page one of Google for that very query, so the omission looks like a glitch. It rarely is, and the cause can usually be found in an afternoon.

Below are seven causes. Six rest on our own citation research: a 500 query study across Google AI Overview, ChatGPT and Perplexity, and an August 2026 study of 6,601 AI answers. The freshness cause rests on third party studies we have summarised. After the seven causes comes a 30 minute diagnosis to find which one applies to you, what changes when the question names Singapore, and a 30 day plan.

1. Their page answers the exact question; yours covers many

Most company sites were built for an older Google reward: one broad service page or pillar guide ranking for a cluster of keywords. AI engines answer one question at a time and pick the page that fits it.

In our one deep page per question study, we recorded every URL that Google AI Overview, ChatGPT and Perplexity cited across 500 queries in April 2026, 6,083 unique pages in all. Of those, 5,451 pages, or 89.6%, were cited for exactly one query. No page was cited for more than 8 of the 500. CNET's best VPN guide, one of the most cited pages in the whole study, reached 8 queries.

Here is how that plays out. A competitor publishes a page titled "How much does a website redesign cost in Singapore?" You have a Services page with one paragraph on pricing. When a buyer asks the cost question, the engine wants the page built to answer it.

Splitting into hundreds of thin pages fails too. The median cited page in the same study carried 16 headings, 14 lists and 32 external links. A template page with a flat structure and no references resembles nothing in the winners' table.

Check: write down the five questions buyers ask before they shortlist a supplier like you. For each one, can you point to a single URL whose whole job is to answer it?

2. Other websites talk about them and rarely mention you

An AI engine has two ways to name a company. It can read that company's page while it builds the answer, or it can name the company because the wider web already treats it as part of the category. Our study of being quoted versus being known by AI separated the two across 6,601 answers from ChatGPT, Google AI Overview and Perplexity, collected daily through August 2026.

Vendor A was the most quoted company in the dataset, with its website cited 8,154 times. Take its own site out of the answers and only 5.6% of its mentions survive. Vendor C was cited 479 times, roughly a seventeenth as often, yet it was named 1,226 times, and 74.4% of those mentions arrived with no citation to its own site. The gap comes down to other people: 141 pages across 49 other websites mention Vendor C. Vendor B had none.

Outside data agrees. SE Ranking's analysis of 129,000 domains, which we summarised in what drives AI citations, found that the number of separate websites linking to a site, its referring domains, was the strongest predictor of ChatGPT citations: sites with more than 32,000 referring domains were 3.5x more likely to be cited. A 2026 analysis by AirOps and Nobori, covered in the same summary, found that roughly 85% of AI visibility does not come from a brand's own website.

If your competitor appears in Singapore "best of" roundups, comparison articles, review platforms and community threads, and your brand appears mainly on your own site, this is probably your biggest cause, and the slowest to fix.

3. You optimised for one engine; they show up in the one being asked

If your agency reports a single AI visibility score, ask which engine it measures. In the study behind AI search being three separate channels, 94.2% of the 6,083 cited URLs appeared in exactly one engine. Not one URL was cited by all three, and ChatGPT and Google AI Overview shared zero citations across 500 identical queries.

The engines also disagree on when to cite at all. In that automated run, where ChatGPT's web search was honoured only intermittently, ChatGPT returned zero citations for informational and YMYL (health and money) queries and answered them from its own training. Interactive sessions may behave differently. Google AI Overview cited sources on 96 of every 100 informational queries and only 10 of every 100 local ones, where Perplexity cited on all 100.

So a competitor can be missing from Google AI Overviews and still own ChatGPT's answer to a commercial question. The study ran in US English in April 2026, so treat the exact rates as a dated snapshot; the three way split is the durable part.

Check: run each priority question separately in ChatGPT, Google AI Overviews and Perplexity, and record the result per engine.

4. Their content is newer

AI answers lean heavily on recent material. The AirOps 2025 research referenced in our citation factors summary found that 95% of ChatGPT citations come from content published in the last 10 months. Digitaloft found that 76.4% of the most cited pages had been updated in the last 30 days.

A competitor that refreshes its pricing and comparison pages every quarter keeps landing back inside that window. A service page last edited in 2023 sits outside it, however well it ranks on Google. The update has to be real: revise the figures, examples and prices, then say at the top what changed. A new date on old content gives the reader nothing.

Check: note the visible dates on the pages AI cited for your competitors, then on the page you expected to be cited.

5. AI can tell who they are; it can't tell who you are

Before an engine can recommend a company, it has to work out who a page belongs to and what that company does. In our research on how each AI engine treats schema markup, 80.8% of the pages ChatGPT cited carried schema markup, against 59.5% for Perplexity and 48.4% for Google AI Overview. The leading types on ChatGPT's cited pages were Organization (68.1%), BreadcrumbList (52.7%) and Article (40.1%). Together they declare who published the page and where it sits.

Two caveats apply. ChatGPT's figure rests on 182 pages, so read it as directional. And a controlled Ahrefs test, discussed in the same study, found no major uplift when 1,885 pages added schema markup, measured against 4,000 matched controls. Schema describes the well maintained sites engines tend to pick; adding it does not buy a citation. FAQ schema, the most promoted tactic of the last two years, separated nothing: all three engines cited pages with FAQ markup at about 9%.

The useful work is making your identity unambiguous, in visible text first and code second. A searchVIU test covered in the same study found that five AI systems, ChatGPT and Google AI Mode among them, read the visible HTML; none returned a price that existed only in markup. So state plainly on your homepage and About page what you do, for whom and where, for example "payroll software for Singapore SMEs". Use one company name everywhere, add Organization schema whose sameAs field lists your official profiles, such as LinkedIn, and keep the same description on your directory listings.

Check: ask each engine "What does [your company] do, and who is it for?" A vague or wrong answer means the engine cannot place you in the category, so it names a competitor it can.

6. Your answer is in the page but not where AI looks

Teams usually work on headlines: the question goes into the H1 and the title tag. Our study of what actually earns an AI citation examined 39,284 pairings of a question with a page that ChatGPT, Google AI Overview or Perplexity cited for it in August 2026. The typical cited page shares only a quarter of the question's words in its headline, and 8% share none. The body of a cited page covers 80% of the question's wording.

The engines read the whole page and look for a passage that resolves the question. The result that surprised us came from comparing 135 home pages that were quoted against 2,703 that never were, all on websites the engines quoted elsewhere. The ignored pages were longer, carried more headings and list items, and were more likely to include structured markup. A page that covers fifteen subjects holds a weaker claim to any one of them.

If your answer to "how long does implementation take?" sits in paragraph nine of a general product page, a competitor who answers it in two direct sentences under its own heading is the easier source to quote.

Check: find the sentence on your site that answers each priority question, and move it to the start of a section headed by that question.

7. Ranking on Google is not the same as being cited

This is the cause behind the familiar complaint: "We rank, so why are we invisible?" In our study of the pages that win both Google rankings and AI citations, only 985 of 6,576 pages, 15%, both ranked in Google's organic top 10 and were cited by an AI engine for the same query. Another 1,684 ranked in the top 10 with no AI citation at all, and 3,907 were cited without ranking in the top 10.

The pages that won both read like research documents: the median one carried 56 external links, roughly twice the 28 to 29 of pages that won only one system. Local intent is the exception: only 5% of the pages that won both came from local queries, because Google answers local questions with its map of business listings while AI engines cite directories.

The overlap is also shrinking. Ahrefs' March 2026 study of search results for 863,000 keywords, cited in the same research, found that only 38% of pages cited in Google's AI Overviews still rank in the top 10, down from about 76% in mid 2025. A page one ranking is a good start. It does not make your page the one the engine quotes.

Check: put your ranking page next to the page AI cited instead, and compare sources linked, scope and how quickly each reaches the answer.

8. A 30 minute diagnosis

You can narrow down the cause in about half an hour with a browser and a spreadsheet. Start with access, because a blocked crawler makes every other fix irrelevant.

Minutes 1 to 5, crawler access. Open yourdomain.com/robots.txt and look for rules naming GPTBot, OAI-SearchBot or PerplexityBot, and for a blanket Disallow that catches them. OpenAI's crawler documentation states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers, and that GPTBot, its training crawler, is controlled separately. Perplexity's crawler documentation describes PerplexityBot as the crawler that surfaces and links websites in Perplexity's search results.

On Cloudflare, check the dashboard too, because a block applied there does not appear in robots.txt. Cloudflare announced on 1 July 2025 that it was the first infrastructure provider to block AI crawlers accessing content without permission by default, with every new domain asked upfront whether to allow them. Its Block AI Bots documentation adds that from 15 September 2026, new domains block bots classified as Training or Agent on pages that display ads, with Search crawlers still allowed. The controls sit under Security Settings, Configure AI bot policies.

Minutes 5 to 30, symptoms. Run three to five of your buyer questions in each engine and match what you see to the table.

What you seeLikely causeHow to check
You are missing from every engine, even for questions about your own brandCrawler access (check this first)Read robots.txt and your Cloudflare AI bot settings, as described above the table
The cited competitor has a page dedicated to the exact question; you do not1. No dedicated pageSearch your own site for the question and name the one URL built to answer it
The engine names the competitor without linking to its site2. Third party mentionsCount roundups, review sites, Reddit and Quora threads that name each brand in your category
You appear in one engine and vanish in another3. Engine mismatchRun the same prompt in ChatGPT, Google AI Overviews and Perplexity and log each result
The cited competitor pages were updated recently; yours were not4. FreshnessCompare visible updated dates on the cited pages and on yours
The engine describes your company vaguely or wrongly5. Entity clarityAsk each engine what your company does and for whom; check your Organization schema
You cover the answer, but deep inside a long page6. Answer placementFind the answering sentence and check whether it opens a section under a matching heading
You rank in Google's top 10 and are still not cited7. Ranking versus citationCompare sources, scope and depth of your page against the cited page
Symptom to cause map for the seven causes above, plus crawler access. Run the checks in the order shown; the first row takes five minutes, and a block found there explains every other symptom.

Doing this by hand works for a handful of questions. For a larger set, our free AI visibility tracker checks how ChatGPT, Gemini, Perplexity and Google AI Overview cite your brand across more than 50 real prompts.

9. What changes for Singapore questions

Our own tracking shows which sources the engines reach for when a buyer's question names Singapore. In 1,588 answers from ChatGPT, Google AI Overviews and Google AI Mode to 19 Singapore agency buying questions (5 September to 4 October 2026, tracking by Novastacks), reddit.com was cited in 76 answers and quora.com in 71.

For a Singapore brand, that puts community threads inside the citation pool: Reddit appeared in about one answer in 21 and Quora in about one in 22. Two further steps are our recommendation rather than something this tracking measured. If a Singapore "best providers" list exists for your category, get onto it, because a list that leaves you out cannot name you when an engine cites it. And give Singapore buyers a service page scoped to Singapore, with local pricing, clients and terms.

We tracked agency buying questions only. Other categories will draw on different local sources, so run your own Singapore question set before you generalise from ours.

10. A 30 day plan

If the diagnosis turned up a crawler block, fix that on day one. Then work through the weeks in order.

Week 1, baseline. Pick 15 to 25 questions your buyers actually ask, in their words, with Singapore versions where location matters. Run each one in ChatGPT, Google AI Overviews, Google AI Mode and Perplexity. Record which brands each answer names and which URLs it cites. This sheet is your day 30 benchmark.

Week 2, one deep page per priority question. From the baseline, choose the three to five questions where a competitor is cited with a dedicated page and you have none. Build one page for each, scoped to that question: the answer in the first hundred words, headings that follow the buyer's next questions, real figures, and links to your sources.

Week 3, entity, schema and freshness. Rewrite the homepage and About page so they state what you do, for whom and where. Add Organization, BreadcrumbList and Article schema with consistent names and sameAs links. Refresh your most important buyer pages with current figures and a visible updated date.

Week 4, third party mentions. List the roundups, comparison articles and directories the engines cited in week 1, and ask to be included, with evidence that earns the place. Answer real questions on Reddit and Quora as a named, disclosed member of your team. Our guide to Reddit SEO and AI citations covers which threads engines quote, and our Quora SEO guide covers the same ground for Quora answers.

At day 30, rerun the week 1 questions and compare per engine, then keep a monthly rerun going. If you would rather have this cycle of baseline, build and measurement run for you, that work is answer engine optimisation (AEO), and it is how our AEO agency service starts every engagement.

Frequently asked questions

Why does ChatGPT recommend my competitors instead of my company?

Usually because a competitor has a page dedicated to the exact question, is named on more third party websites, or has fresher content. In our research, 89.6% of the pages AI engines cited won a citation for exactly one query, and a vendor named on 141 pages across 49 other websites earned 74.4% of its mentions without its own site being cited.

If I rank on page one of Google, why am I not in Google AI Overviews?

Ranking and citation overlap far less than most teams expect. Of 6,576 pages in our study, only 15% both ranked in Google's top 10 and were cited by an AI engine for the same query. Ahrefs found in March 2026 that only 38% of pages cited in AI Overviews still rank in the top 10.

Does adding schema markup get my brand cited by AI?

Not by itself. In our study 80.8% of the pages ChatGPT cited carried schema, against 48.4% for Google AI Overview, but a controlled Ahrefs test of 1,885 pages found no major uplift from adding schema markup. Treat Organization, BreadcrumbList and Article schema as part of a clear identity, and do not count on FAQ schema, which all three engines cited at the same rate of about 9%.

How do I check if my website blocks AI crawlers?

Open your robots.txt file and look for rules covering GPTBot, OAI-SearchBot and PerplexityBot. OpenAI states that sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. If your site uses Cloudflare, also check its AI bot settings. Since 1 July 2025 Cloudflare has asked every new domain upfront whether to allow AI crawlers, and from 15 September 2026 its default for new domains blocks Training and Agent bots on pages that display ads while still allowing Search crawlers.

Do I need to optimise for every AI engine?

Not equally. In our 500 query study, 94.2% of cited URLs appeared in exactly one engine, and ChatGPT and Google AI Overview shared no citations. Find out which engine your buyers use and which one already sends you referral traffic, and start there.

Do Reddit and Quora matter for AI visibility in Singapore?

They are part of the citation pool. In 1,588 answers from ChatGPT, Google AI Overviews and Google AI Mode to 19 Singapore agency buying questions between 5 September and 4 October 2026, reddit.com was cited in 76 answers and quora.com in 71.

About the author

I am Eki Riandra, Head of AEO at Novastacks, an AI-driven marketing agency specialising in AI search. I am a Google Product Expert in Search Central, spoke at Search Central Live Jakarta 2024, and my Core Web Vitals work at Tokopedia became a Google case study. Across our principals we hold more than 20 years combined in growth marketing and search. For this article I reread six Novastacks citation studies and our summary of SE Ranking's citation factor data, pulled the Singapore figures from our own prompt tracking, and checked the OpenAI, Perplexity and Cloudflare crawler documentation on 4 October 2026.

Sources

  1. Novastacks, "One Deep Page per Question". 500 queries across five intent categories, 6,083 unique cited URLs, US English, April 2026. 89.6% of cited pages won exactly one query; no page won more than 8.
  2. Novastacks, "Being Quoted Is Not the Same as Being Known". 89 questions, 6,601 answers and 61,536 citations from ChatGPT, Google AI Overview and Perplexity, August 2026. Vendor A 8,154 citations and 5.6% earned recognition; Vendor C 479 citations, 1,226 mentions, 74.4% earned recognition, named on 141 pages across 49 websites.
  3. Novastacks, "What Drives AI Citations? Data from 129,000 Domains". Summary of SE Ranking's 129,000 domain analysis (referring domains strongest predictor; 32,000+ referring domains 3.5x more likely to be cited), AirOps 2025 (95% of ChatGPT citations from content published in the last 10 months), Digitaloft (76.4% of most cited pages updated in the last 30 days), and AirOps with Nobori (about 85% of AI visibility off a brand's own site).
  4. Novastacks, "AI Search Isn't One Channel. It's Three.". Same 500 query dataset. 94.2% of cited URLs in exactly one engine; zero in all three; ChatGPT and Google AI Overview shared zero citations; citation coverage by intent type.
  5. Novastacks, "Every AI Engine Plays by a Different Rulebook". 5,090 cited pages with HTTP 200. Schema on 80.8% of ChatGPT's cited pages (182 pages), 59.5% Perplexity, 48.4% Google AI Overview; FAQPage about 9% on all three. Discusses the Ahrefs controlled schema test (1,885 pages, 4,000 controls) and the searchVIU markup test.
  6. Novastacks, "What Actually Earns an AI Citation". 39,284 question and page pairings, August 2026. Headlines share a quarter of the question's words, 8% none; body covers 80%; 135 quoted versus 2,703 never quoted home pages.
  7. Novastacks, "The Pages That Win Everywhere". 6,576 pages in the analysis sample. 985 (15%) ranked in Google's top 10 and were cited by AI; median 56 external links versus 28 to 29; local 5%. Cites Ahrefs' March 2026 study of search results for 863,000 keywords (38% of AI Overview citations rank top 10, down from about 76%).
  8. OpenAI, crawler documentation (OAI-SearchBot and GPTBot). Checked 4 October 2026. Sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers; OAI-SearchBot and GPTBot settings are independent.
  9. Perplexity, crawler documentation. Checked 4 October 2026. PerplexityBot is designed to surface and link websites in search results on Perplexity.
  10. Cloudflare press release, "Cloudflare Just Changed How AI Crawlers Scrape the Internet-at-Large". 1 July 2025. First infrastructure provider to block AI crawlers accessing content without permission or compensation by default; every new domain asked whether to allow AI crawlers.
  11. Cloudflare Docs, "Block AI Bots". Last updated 1 July 2026, checked 4 October 2026. From 15 September 2026 new domains block Training and Agent bots on pages that display ads, Search allowed; mixed purpose crawlers blocked by all AI training block settings.
  12. Novastacks prompt tracking, Singapore agency buying questions. 1,588 answers from ChatGPT, Google AI Overviews and Google AI Mode to 19 questions, 5 September to 4 October 2026. reddit.com cited in 76 answers, quora.com in 71. Reported at category level; no brand named.

Find out why AI names your competitors

Get a free AI visibility audit: we run real buyer prompts through ChatGPT and Perplexity, show who they recommend instead of you, check whether Google's AI Overview appears for your keywords, and list your three highest impact fixes.