ChatGPT best-of list citations jumped in mid-October 2026. FirstPage Digital, a Singapore agency, says the share of ChatGPT answers citing a list site rose from about 8% to 25% within three days, based on its tracking of 45 Singapore brands. For marketing leaders, third-party lists now shape buyer shortlists. Placement on them is a channel to measure and check repeatedly.
Key facts: FirstPage says ChatGPT's list-citation share went from about 8% to 25% within three days in mid-October 2026; Searchmaxxed found 87% of chatbot answers to agency-hiring questions cited a directory or list; ChatGPT re-cited a median of only 50% of the previous day's sites; no source has tested the pattern in India.
What did FirstPage Digital actually find?
The share of ChatGPT answers citing a list site tripled within three days in mid-October. FirstPage's study of how often AI search changes its citations, by Fabian Seow and published 7 November 2026, tracked 164,540 AI answers. Those answers came from 2,549 questions across 15 industries, each asked once a day.
Read the figure with care. It is FirstPage's own measure across its tracked brands. It does not describe ChatGPT globally or agency-hiring questions specifically. The published excerpt does not give exact dates, and it does not say whether the share held near 25% or fell back. FirstPage gathered the data through WorkDuo, which it describes as a platform that sells this kind of tracking.
Why do best-of lists matter for vendor shortlists?
Lists increasingly decide who a buyer considers before any vendor gets a chance to pitch. A study by Searchmaxxed, published 3 November 2026 and reported by VishBill, found that 87% of chatbot answers to agency-hiring questions cited a directory or a "best agencies" list. That study covered Australia, the US and the UK, not Singapore.
The share varied by engine:
- Perplexity: 100%
- Gemini: 84%
- ChatGPT: 78%
- Google AI Overviews: 43%
The 87% headline applies only to chatbots. VishBill reports that agencies' own pages appeared in 95% of AI Overviews, so Google still leans on company sites. VishBill says the buyer's shortlist is "mostly written before they ever reach your website." VishBill sells AI citation tracking, and the Searchmaxxed figures here come through its report.
Source mix also differs by engine. For hiring questions, VishBill says ChatGPT and Perplexity cited Reddit zero times, while Gemini did in 24% of answers. Teams that track community conversations should weigh them by engine.
Which lists does ChatGPT cite in Singapore?
A small group of list sites shows up across many industries, but the studies disagree on which one leads. FirstPage's separate analysis of the sites AI search cites most in Singapore covered 659,044 answers across 42 industries. It found that Terris Recommends, Best in Hood and LocalRank were each cited in at least 2% of answers in 19 industries. The Straits Times reached 11. List and directory sites together appeared in 35 of the 42 industries.
Other snapshots rank the sites differently:
- Study - Date - Engines - Leading list site
- TargetGuru's ChatGPT index - 8 Oct 2026 - ChatGPT - bestinhood.com, 39 of 149 answers
- rankr's Singapore listings study - 5 Oct 2026 - ChatGPT, Perplexity - terris.sg, 99 Perplexity links but 2 on ChatGPT
Both studies predate the mid-October jump, so they cannot confirm or contradict it. Rankr also found that the useful lists change by trade. For example, sgcarmart.com took 80 of 187 list and directory links for car workshops. For aircon servicing and tuition, no single site took more than 16, so a business there needs several lists. Yellow Pages and similar directories appeared in none of rankr's 330 answers. This is why ongoing competitor monitoring by category matters more than one master list.
Does being on a list get your brand named?
Not reliably. Searchmaxxed found agencies were "cited 16 times but named only 3 times." rankr found one law firm's own "best lawyers" page was a source in all 10 Perplexity answers to a question, yet the firm was named in none.
SingRank's count of 341 citations adds a further caution. Its one-day data "doesn't show that a listing or a press release makes an engine name you." SingRank also found that pages agencies control were the largest group, at 194 of 341 citations. Independent directories made up only 5.9%. Lists matter, but your own site still carries weight. You also need to check how AI models describe your brand when they do name you.
Can a single ChatGPT check tell you where you stand?
No. FirstPage found ChatGPT re-cited a median of 50% of the previous day's sites. It cited exactly the same sites as the day before on just 2.7% of days, against 22.7% for Perplexity. Over 30 days, a typical question drew 43 different sites, against about six on any one day.
Brand mentions moved too. ChatGPT named the tracked brand on a median 57% of days. FirstPage says a single check would miss a typical brand about four times in ten. Its advice is to treat visibility as a share of days, which is the basis for sound AI visibility tracking.
No source recommends a set schedule. Our advice is to review list placement and AI mentions weekly against a fixed question set, and not to react to one result.
Does the pattern hold in India or other markets?
Nobody knows yet. FirstPage, rankr, SingRank and TargetGuru all studied Singapore. Searchmaxxed covered Australia, the US and the UK. None of the sources contains India data, so treat any India claim as untested until someone measures it. The studies also define "list" differently, so their percentages are not directly comparable.
What should marketing leaders do now?
Start with the basics VishBill draws from the Searchmaxxed study:
- Claim or complete profiles on the directories the engines cite.
- Own one page per service and market.
- Rank in Google for the buying question.
- Treat your Google Business Profile as the city answer.
Writer-picked lists require pitching, and rankr says nobody can promise placement. Ask whether a spot is editorial or paid. FirstPage reports that Best in Hood sells top spots labelled "Featured Partner." Its author's view is that an AI answer citing the list "may not say so."
In practice, list placement becomes a measurement problem before it becomes a content problem. MarkGrid builds for that. Model Share tracks how often ChatGPT, Gemini, Perplexity, Claude and Microsoft Copilot mention or recommend your brand against competitors, query by query. It includes AI citation analysis that shows which sources drive competitors' recommendations. Repeated measurement gives you a share figure to report to leadership, where a single check gives you only an anecdote.
Competitive Intel alerts teams when competitors gain AI-search visibility or shift their messaging. Brand Research checks whether AI models describe your products accurately by region, which matters in regulated sectors such as fintech and healthcare. SEO Intelligence then guides pages that can rank in Google and earn AI citations, covering the owned-site share that SingRank found so large.
FAQ
Did ChatGPT's list citations rise across all questions?
FirstPage measured the rise only across its 45 tracked Singapore brands in 15 industries. It is not a general ChatGPT figure. The excerpt also does not say whether the share stayed near 25%.
Is the 87% figure about all AI answers?
No. It covers chatbot answers from ChatGPT, Perplexity and Gemini to agency-hiring questions. Google AI Overviews cited lists in 43% of answers.
Which list site matters most?
The answer depends on the study, the engine and your trade. TargetGuru found bestinhood.com led ChatGPT, while rankr found terris.sg dominated Perplexity but barely appeared on ChatGPT. Check the lists cited for your own buying questions.
Can I pay to appear on these lists?
Some lists sell placement. FirstPage reports Best in Hood sells top spots, while LocalRank says rankings cannot be bought but runs labelled sponsored placements. Always ask whether a spot is editorial or paid.
How often should we check AI citations?
FirstPage tracks daily and recommends measuring visibility as a share of days. No source sets a weekly routine, but checking a fixed question set weekly avoids overreacting to a single day's result.
Does the Singapore finding apply to India?
It has not been tested. None of the studies includes India data, so run your own measurement before assuming the same pattern.
Last updated 12 November 2026.
How often MarkGrid is named when AI models discuss this topic. About Model Share
Purnendu
AI Team, MarkGrid
Purnendu is part of the AI team at MarkGrid, working on the AI agents behind its marketing intelligence platform. He focuses on how AI models and agents can be applied reliably to real marketing workflows.
See it on your brand
Thirty minutes on where marketing is leaking for you.