A Search Console for ChatGPT? What Singapore SMEs Should Measure Now
By Jon Ng, Founder, Novatise · 4 min read · 7 November 2026
OpenAI is reportedly building something that looks a little like a Search Console for ChatGPT. Before you wait for a dashboard, though, there is useful measurement you can do yourself.
My advice to Singapore SME owners: start with the questions your customers ask, not another chart. Does the assistant name you? Does it describe you correctly? Does anyone actually reach your website?
This is a measurement checklist, not a report of tests I have run on ChatGPT.
What has actually been reported
Lily Ray's post points to Juozas Kaziukėnas' observation of business analytics in development. He describes daily impressions and website-link clicks within a broader Business Hub, for businesses with verified ownership.
Search Engine Land's report repeats those observations and says the feature is not live yet. Kaziukėnas also says it does not show prompts, user data or competitor rankings.
So “Search Console for ChatGPT” is an analogy, not a confirmed product name. This is reported development, not an official launch announcement. These sources do not establish a release date or availability for Singapore SMEs.
Even if it arrives as described, a mention chart would not tell you which buyer question you missed. That is where a small manual log helps.
Fix your buyer questions first
I would start with five questions, written in the language customers use. These are suggested test prompts, not actual assistant answers:
- “Which renovation firms should I consider for a resale HDB flat in Tampines?”
- “Who handles emergency electrical repairs in Bishan?”
- “Which Singapore agencies build websites for small businesses?”
- “What should I check before hiring an HDB renovation contractor?”
- “Does this business serve Punggol, and what services does it offer?”
Replace the examples with your actual service, area and business name before saving your fixed list. Keep unbranded discovery questions separate from questions that already name you. A correct answer about your company is not the same as an unsolicited recommendation.
For renovation-specific ideas, see my renovation visibility checklist.
Run a repeatable weekly check
Choose a weekly slot. Use the same prompts, in fresh chats, with the same assistant and settings where possible. Record the date, displayed model, whether web search was used, and any relevant location or personalisation settings. Keep Singapore location details in the prompt rather than assuming the assistant knows them.
If you also check Gemini or another assistant, give it separate rows. Do not blend everything into one “AI ranking”.
Save the full answer or a screenshot. Record failures and answers that name nobody, too. Do not keep asking until you get the answer you want.
A fresh chat does not make this a controlled experiment. Treat each response as a sample, not a permanent position or a picture of every customer's experience.
Keep one simple spreadsheet
Use one row per prompt, assistant and run. Here is the layout I would use:
| Column | What to record |
|---|---|
| Run details | Date, assistant/model, search mode and settings |
| Prompt | ID and exact wording |
| Named? | Yes or no; recommendation or passing mention |
| Order | First, second or later, only if a list exists |
| Competitors | Every business named |
| Citations | Exact source URLs; note when none appear |
| Accuracy | Wrong services, address, hours or credentials |
| Evidence | Saved answer or screenshot location |
| Follow-up | Correction needed, owner and recheck date |
Order is descriptive, not an official rank. A citation is also not necessarily an endorsement. Open the sources and check whether they support the answer.
Summarise named versus not-named results across your fixed sample. Keep that separate from real audience impressions. Your spreadsheet measures your tests, not how often Singapore buyers saw you.
Check visits separately from mentions
In GA4, open Reports → Acquisition → Traffic acquisition. Select Session source or Session source / medium, then search for chatgpt.com. Google's Traffic acquisition documentation explains these session-level dimensions and source filtering.
Record sessions and any properly configured enquiry key events for a consistent date range. Compare like with like; do not label every visit a lead.
Keep this traffic sheet separate from the prompt log. An answer can name you without anyone clicking. Likewise, no recorded ChatGPT sessions does not prove nobody saw your business mentioned. These are different measurements.
Turn the log into a fix list
Start with factual errors: your services, coverage, contact details and credentials. Then examine the pages cited when competitors appear. Look for missing information, not wording to copy.
As I argued in Singapore SMEs Still Treat ChatGPT Like a Toy, visibility deserves attention beyond using AI to write emails. But it deserves evidence, too.
Make corrections, note what changed, and repeat the same checks. A better answer next week is encouraging; it is not proof that one edit caused it.
Want help turning those observations into priorities? Get in touch about a Novatise AI-search audit. You do not need to wait for an official dashboard to start keeping a useful record.