Asked if ChatGPT can find a website, most owners check Google instead. On 25 September 2026 we measured the eight Singapore company sites ranking on Google page one for "AI search optimization agency Singapore": seven answered a normal fetch, only two publish FAQ schema, and two publish no structured data at all.
If a buyer asks an AI assistant for a recommendation in your category, the assistant quotes pages it could fetch and understand. Below is the same five-point check we ran on those eight sites. It takes about five minutes and it needs no account.
yourdomain.com/robots.txt
Look for lines naming GPTBot, PerplexityBot, ClaudeBot or Google-Extended followed by Disallow: /. That line tells the crawler to stay out, and nothing else on this list can fix it. In our sample, none of the sites we could read blocked those crawlers.
FAQPage
That is structured data labelling a question and its answer. Two of the seven sites we could measure had it — mediaplus.com.sg and awebstar.com.sg. Two had no structured data at all: qoulomb.com and exabytes.sg.
Our eight-site sample served a median 233 KB of HTML, from 85 KB to 943 KB. Every extra hundred kilobytes is more for a crawler to fetch and parse before it reaches the sentence that answers the buyer's question. Fix the weight first; editing copy on a heavy page fixes the wrong thing.
Answer engines quote passages, not pages. A quotable passage answers one buyer question in 40 to 60 words and carries a number, a date or a fact — for example, what your service costs, how long it takes, or how many customers you served last month.
yourdomain.com/sitemap.xml and check the addresses are real pages
Until 25 September 2026, this site's own sitemap listed /#pricing and /#audit — anchor links, which give a crawler no page to fetch. It now lists four real addresses. If your sitemap points at anchors, an engine sees one page where you thought you had five.
What this check does not tell you. It tells you whether an engine can read you and quote you — not whether it will. No agency can promise a position in ChatGPT or Google AI answers, including us. What is measurable is the floor: crawler access, structured data, page weight and quotable passages. Those four are what this check covers, and they are the four we work on.
Asked if ChatGPT can find a website, most owners check Google instead. On 25 September 2026 we measured the eight Singapore company sites ranking on Google page one for "AI search optimization agency Singapore": seven answered a normal fetch, only two publish FAQ schema, and two publish no structured data at all.
Open yourdomain.com/robots.txt in a browser and look for lines naming GPTBot, PerplexityBot, ClaudeBot or Google-Extended with Disallow: /. In our 25 September 2026 sample of eight Singapore sites, none blocked those crawlers. If a crawler is blocked, nothing else on this list matters.
FAQ schema is structured data that labels a question and its answer in your page code. It matters because answer engines quote passages, and a labelled question-answer pair is unambiguous. In the same sample, only two of the seven sites we could measure published FAQPage schema; two published no structured data at all.
It affects whether a crawler gets to your words. The eight Singapore sites we measured on 25 September 2026 served a median 233 KB of HTML, from 85 KB to 943 KB. Every extra 100 KB is more to fetch and parse before the crawler reaches the sentence that answers the buyer's question.
Nobody can give a date, because the engine decides. What can be fixed today is reachability and structure. In practice the work order is: unblock crawlers, add labelled question-answer pairs, cut the page weight, then publish one self-contained answer per buyer question with a real number in it.
Send your website address and we run these five checks on your live pages, plus three competitors, and send back your GAS Score. It is free and there is no card.
Get the free audit See what the paid tiers costRelated: the full nine-site measurement · AIFindr pricing · what AIFindr does.