Sample report: fictional business. Every name, number, result and screenshot reference in this document is made up to show what a weCDEF audit report looks like. "Sample Waterproofing Pte Ltd", its website, Firms A to D, directory A, job-matching directory C, site B and forum C do not exist. The register names are real, but the results shown against them are invented. The prompts are taken from weCDEF's waterproofing prompt set (version 1.0); every result shown against them is invented.

AI visibility audit: Sample Waterproofing Pte Ltd (fictional)

Sample report: fictional business.

Summary

Sample report: fictional business. All figures in this section are made up.

• Readiness score: our own 100-point checklist (not an industry standard) of the things AI tools and search engines can find about your firm, such as your Google profile, reviews, website content, listings, licences, and whether AI crawlers can read your site.
• Outcome: how often ChatGPT and Perplexity named you when we asked each question 3 times ("named in X of 3 runs"), and your share of voice, meaning the share of all firm mentions in those answers that went to you rather than to each competitor.

The two are reported separately: the outcome is not part of the readiness score.

Readiness score: 47 out of 100 (Patchy)

Bands: Strong 70–100 (most of the signals search engines and AI tools use to find, understand and trust a firm are in place), Patchy 45–69 (some key signals are in place, but clear gaps remain), Weak 0–44 (few of those signals are in place). The bands describe what you have in place, not what AI answers currently say. The readiness score is weCDEF's own checklist (rubric v2), not an industry standard or a precise measurement: two firms with the same score can have quite different problems, so read the fix list, not just the number.

Outcome: where you show up in AI answers today

We asked ChatGPT and Perplexity the same 4 questions a homeowner in Singapore might ask when looking for a waterproofing contractor. Each question was asked 3 times on each tool, because answers change from one asking to the next.

Share of voice (all 24 runs, ChatGPT and Perplexity together):

You were named in 6 of 24 runs (fictional), mostly for the toilet waterproofing question (WP-02), which matches the service page you've written most about. Across all runs you had 13% of firm mentions, against 31% for Firm A (fictional). For general questions such as "best waterproofing contractor in singapore", the tools linked to directories and roundups where you aren't listed.

The 3 biggest fixes

1. Fix your Google Business Profile phone number and category (fix 1 in section 11). The phone number differs from your website, and ChatGPT repeated the old one (fictional). The primary category is "Contractor" rather than "Waterproofing service". (Impact: High. Effort: Low.)
2. Get listed on the directory the tools linked most (fix 3). Directory A (fictional) was linked 9 times, and you're not on it. Lifestyle site B's roundup was linked 5 times, but it's editorial, so the most you can do is make sure your details are easy for writers to find. (Impact: High. Effort: Medium.)
3. Give each main service its own page (fix 4). You were named only for toilet waterproofing, the one service with its own page. Roof, wall and leak detection share a single page, with no area served in text. (Impact: High. Effort: High.)

These are taken from the high-impact items in section 11, where the full list is ordered by impact first (High, Medium, Low), then by effort (Low, Medium, High).

What we would measure next

On or around 2 Dec 2026 (about day 30 of a pilot) we would run the same 4 prompts, word for word, 3 times each on the same tools and in the same way, re-score the readiness checklist, and compare:

• the outcome: how many runs name your firm and link to your website ("named in X of 3 runs" for each question), and your share of voice against the firms named alongside you;
• your readiness score and each area's score;
• whether your Google Business Profile phone number and category now match your website, whether you're listed on directory A, and whether your HDB registration is stated in text.

We report the numbers as they come out, including if they stay the same or go down.

1. How to read this report

Sample report: fictional business. All figures in this section are made up.

Readiness status labels (used on every checklist item)

Each readiness area's status follows the items we could check: Good if every checked item is Good, Missing if every checked item is Missing, Not checked if we couldn't check any, otherwise Partial. AI crawler access is the exception: its status follows its score (Good 8–10, Partial 4–7, Missing 0–3), because blocking crawlers used only for AI training doesn't count against you.

How an area is scored

Each item has the points shown in its section. An area's score is the points you earned, divided by the points available on the items we could check, times the area's share of the 100 (for example, 13 of 25 points in Search presence is 5 out of 10). Halves are rounded up, so 2.5 becomes 3. That's why each area shows "scored on X of Y items checked". An item we checked and didn't find is Missing and counts as zero. If we couldn't check any item in an area, that area is left out, the total is re-scaled to 100 over the other areas, and the score is marked provisional. Otherwise the total is simply the sum of the area scores shown.

Outcome: AI answer results

For each question and tool we report "named in X of 3 runs". One run that names you and two that don't is a weaker result than three out of three, and it can change from month to month. A run where the tool gave no answer, or an error, is recorded as such and left out of the count.

Share of voice is the number of times your firm was named, divided by the number of times any firm was named, across all runs. It shows how you compare with the firms the tools named alongside you. With 3 runs per question, small changes from one month to the next can be chance, so we look for changes that hold over more than one re-check.

Impact and effort

Impact is our judgement, based on what Google, OpenAI and Perplexity publish about how their tools find and use information. It is not a measured effect, and no single fix will necessarily change what an AI tool says.

2. Outcome: what AI tools tell homeowners

Sample report: fictional business. All figures in this section are made up.

Why this matters: some homeowners now ask an AI tool something like "which company is good for toilet waterproofing in singapore? need someone reliable" instead of scrolling through search results. The tool gives a short answer and names a handful of firms. If you're not among them, that homeowner may not get as far as your website.

2.1 The prompt set

We use a fixed list of prompts for each trade. Every prompt names Singapore. The list is saved, versioned and reused every month, so results from one month can be compared with the next. We don't rewrite the prompts to make a firm look better or worse, and we paste them unchanged on every run.

Prompt set WP (waterproofing), version 1.0:

2.2 Results by question

2.3 Share of voice by question

2.4 Run log

Every run is logged with the details below, and every run has a screenshot of the full answer, including its list of sources. The screenshot is the record: if anything in this report disagrees with a screenshot, the screenshot is right.

"Model shown" is the model name the tool displayed on screen, if it showed one. If it didn't, we write "not shown" rather than guess.

2.5 What the answers said about you

• WP-02, ChatGPT (CG-02-r1): named you as "Sample Waterproofing, which offers no-hacking toilet leak repair" (fictional) but gave no link. The phone number it listed is the old one still on your Google Business Profile.
• WP-02, Perplexity (PX-02-r1, PX-02-r3): named you and linked to your "toilet leak repair without hacking" page. This is the only page of yours any tool linked to.
• WP-06 (roof cost), Perplexity (PX-06-r2): named you once, without a price. The price ranges in every WP-06 answer came from competitors' pages; your site doesn't state a price range in text.
• WP-01 and WP-12: the tools linked to directory listings, "best of" roundups and general guides. You don't appear in those sources.

2.6 Sources the tools linked to

The websites linked most often across all runs. They show which pages the tools pointed homeowners to for your trade, and so where it helps to be listed accurately.

3. Search presence (readiness): 5 / 10 (Partial)

Sample report: fictional business. All figures in this section are made up.

Why this matters: Google, OpenAI and Perplexity each describe their AI search features as surfacing and linking to websites. If your website, or a page that names you, comes up near the top for a homeowner's search, there's something for those tools to find.

We ran 3 typical customer searches through our web search tool and looked at the top 5 results. This is a stand-in for what AI tools find when they search, not a full Google results page.

Scoring: 8 points per search if your own site is in the top 5; 4 if you're named on someone else's page in the top 5 (a "best of" list, a directory, a map listing); 0 if you don't appear; 1 if the search tool's summary names you. Area score: 13 of 25 points, scored on 4 of 4 items checked, which gives 5 out of 10.

The search tool's summary named you for the toilet-leak search (fictional).

4. Reviews (readiness): 12 / 20 (Partial)

Sample report: fictional business. All figures in this section are made up.

Why this matters: reviews are one of the few places a stranger describes your work. AI answers can summarise them ("known for punctual service", "some complaints about delays"), and homeowners read them before calling.

Source of review figures: your public Google Maps listing (fictional), as of 2 Nov 2026. If these figures come from a third-party snapshot rather than Google Maps itself, they can lag by weeks; if your Google page shows different figures, tell us and we'll correct them.

Also noted (not scored): owner replies to recent reviews: 3 of the last 20 (fictional). Themes in recent reviews: toilet leaks fixed without hacking; clear quotes; two mentions of slow WhatsApp replies (fictional).

Scoring: rating 4.8 or above = 8, 4.5–4.79 = 6, 4.0–4.49 = 3, below 4.0 or not verifiable = 0. Count 200 or more = 12, 100–199 = 10, 50–99 = 7, 20–49 = 4, 1–19 = 2, none or not verifiable = 0. Recency: 5 if we saw a review dated within the last 6 months. Area score: 15 of 25 points, scored on 3 of 3 items checked, which gives 12 out of 20.

Your rating is good, but only 5 reviews in the past year (fictional) makes the profile look quieter than your workload probably is. A short, polite request to every customer after each finished job, with a direct link to your review page, is a simple place to start. Replying to every review, including the two about slow replies, shows you're active.

How we approach reviews: we suggest ways to ask real customers for honest reviews after a finished job, and to reply to them. Ask every customer, not only the happy ones, and don't suggest a rating or what to write. We never write, buy or swap reviews, and we don't recommend offering gifts, discounts or anything else in exchange for them. Google's policy bans fake and incentivised reviews (including payment, discounts, free goods or services), and asking only happy customers for reviews (Google Maps User Generated Content Policy, "Prohibited & restricted content"). Fake reviews also mislead customers. In Singapore, the Competition and Consumer Commission of Singapore (CCS) treats a business posting fake reviews as an unfair trade practice, and has taken action against businesses that bought them.

5. Structured data and FAQs (readiness): 0 / 10 (Missing)

Sample report: fictional business. All figures in this section are made up.

Why this matters: structured data is a short block of code on your website that states your business facts in a format machines read reliably: your name, trade, phone number, address and services. It doesn't change how the page looks. Google says it uses structured data to understand a page and the businesses it describes, and that the markup must match what the page actually shows. Google also states that no special markup is needed to appear in its AI features, so we treat this as basic housekeeping, not a shortcut. This area also scores whether your site answers real homeowner questions in the page text, because those answers are what a reader, or a tool, can actually use.

We checked the code and text on your homepage and your toilet-leak and roofing service pages.

Scoring: a LocalBusiness block (or a subtype such as RoofingContractor or Electrician) with name, street address and phone = 8; partial (only an Organization block, or no street address) = 4; none = 0. On-site FAQ content (real homeowner questions answered in the page text) = 4, none = 0. Service or Offer markup = 3. Area score: 0 of 15 points, scored on 3 of 3 items checked, which gives 0 out of 10.

Your website builder doesn't add business markup by default (fictional). Adding one block to the homepage is a small job once the phone number in section 7.2 is settled, so both say the same thing.

What good looks like: one clean block on the homepage naming the most specific business type that fits, with details that match your Google Business Profile word for word, and no reviews, ratings or claims in the code that aren't shown on the page. Also noted (not scored): FAQ markup (FAQPage) on the pages we checked: none found. Since August 2023 Google has shown FAQ rich results regularly only for well-known, authoritative government and health websites, so adding FAQ code is unlikely to make your answers appear as drop-downs in Google. That's why we score the visible answers, not the code.

6. Content clarity (readiness): 5 / 20 (Partial)

Sample report: fictional business. All figures in this section are made up.

Why this matters: homeowners ask specific questions: how long does the job take, do you need to hack the tiles, what does it cost, is there a warranty. When an AI tool answers those questions, it can only draw on pages that actually answer them. Google's guidance for its AI features stresses having important content in text and writing helpful, people-first content. A site with one page per service, real price ranges and plain answers, written by someone who does the work, gives them something accurate to use.

Scoring: services pages 5 (partial 2); published prices 5; NAP consistent 5, minor mismatch 2, major mismatch or missing 0. Area score: 4 of 15 points, scored on 3 of 3 items checked, which gives 5 out of 20.

Common homeowner questions your site doesn't answer yet: cost range; how long the job takes; whether tiles need hacking; when the toilet can be used again; warranty length and what it covers; HDB vs condo differences. (On-site FAQ content is scored once, in section 5.)

The answers should come from you: your real price ranges, timings and warranty terms. We can help write them up, but we won't invent figures.

7. Third-party mentions (readiness): 9 / 20 (Partial)

Sample report: fictional business. All figures in this section are made up.

Why this matters: AI tools and search engines find your firm on other sites as well as your own: directories, review platforms, articles and genuine discussions. When the same name, phone number and trade appear everywhere, whichever page they use gives the same facts. When details conflict (an old phone number, a previous company name), an answer can repeat the wrong one.

Scoring: directories and platforms listing you: 3 or more = 7, 2 = 5, 1 = 3, none = 0. Editorial list = 4. Community mentions: positive or neutral = 4, mixed or not verifiable = 2, none or negative = 0. Area score: 7 of 15 points, scored on 3 of 3 items checked, which gives 9 out of 20.

7.1 Listings checklist

Listings for every trade

Directories and platforms for Waterproofing and leak repair

Official registers are covered in section 8. We report only what each platform shows on the date we checked.

Checked on 3 Nov 2026.

7.2 Google Business Profile details (not scored separately)

Your Google Business Profile is the box with your name, map pin, phone number, opening hours and reviews. Google's guidance for its AI features tells businesses to keep it up to date, and it's the first place many homeowners check before they call.

The phone mismatch is the most urgent item. ChatGPT has already repeated the old number (CG-02-r1, fictional).

8. Licences and registrations (readiness): 7 / 10 (Partial)

Sample report: fictional business. All figures in this section are made up.

Why this matters: homeowners often ask trust questions, such as whether a contractor is HDB registered or licensed for electrical work. Official registers are public pages that anyone can check, including AI tools when they search the web. No AI company says it checks these registers, so we can't say how much they affect AI answers. We score them because accurate, checkable licence details help homeowners trust what they read, and a claim that doesn't match the register can do real harm.

Register status: Verified (seen on the official register), Claimed (stated by you, or seen only on a third-party copy of the register), Not verified (we searched the register and you aren't listed, scored as Missing), Not applicable (the register doesn't apply to your work, so it's left out and never marked down). If a register couldn't be reached, the item is Not checked: it's left out of the score and we check it again in a follow-up pass.

• BCA's Contractors Registration System is mainly for firms that tender for public-sector construction work or hire foreign construction workers. For a firm that only works on homes, it's marked Not applicable and never counts against you; if you are registered, it still counts in your favour.
• Electrical and solar work: the EMA licensed electrical worker search (ELISE) lists individual licensed electrical workers, not companies. We check the licensed electrical worker your website names, and report whether that person is listed. If your website names no licensed worker, the item is Not checked until we've asked you, and we recommend showing the worker's name and licence number on your site.

Scoring: main register verified = 4, claimed = 2, not verified = 0. Accreditation or second register verified = 2, claimed = 1, none or expired = 0. Registration names with numbers or workheads in text = 2, logos or names only = 1, nothing = 0. Claims: none contradicted = 2; any claim contradicted by a register (absent, expired, wrong workhead) = 0, and we point it out; register couldn't be reached = Not checked. If an item doesn't apply to your trade, it's left out and the area is scored on the rest. Area score: 7 of 10 points, scored on 4 of 4 items checked, which gives 7 out of 10.

Your HDB listing is the strongest trust signal you have, but it's only shown as a logo image, which search engines and AI tools can't reliably read. Writing it out in text, with your registration details, is a small change. BCA's Contractors Registration System is mainly for firms tendering for public-sector work, so we haven't counted it against you.

Checked on 3 Nov 2026. We report only what each register shows on the date we checked. Whether your firm needs to be on any register is a matter between you and the relevant authority; this report isn't advice on that.

9. AI crawler access (readiness): 9 / 10 (Good)

Sample report: fictional business. All figures in this section are made up.

Why this matters: AI companies send automated visitors ("crawlers") to read websites. AI search tools show and link pages their search crawlers are allowed to read. OpenAI, for example, says: "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers, though can still appear as navigational links." Your site tells crawlers what they may read in a file called robots.txt, and some security settings also turn automated visitors away with a bot check or captcha, sometimes without the owner knowing. Crawlers that only collect pages for AI training don't decide whether you appear in search answers, so blocking them is your choice and isn't marked down. Google-Extended works differently; see the table below.

Finding: The search crawlers are allowed; only GPTBot (used for training) is blocked.

Scoring: if robots.txt blocks Google's or Bing's main crawler from the whole site or your key pages, this area scores 0 and we flag it as urgent. Otherwise: AI search crawlers 3, minus 1 for each one blocked; no bot check 3, a search crawler challenged or refused 1, Google's crawler also challenged (or every crawler) 0; indexing 2, one issue 1, homepage set to noindex or main text missing 0. If you have no website, this area is Missing and scores 0, with the note "no website". Area score: 7 of 8 points, scored on 3 of 3 items checked, which gives 9 out of 10. Status follows the score: Good 8–10, Partial 4–7, Missing 0–3.

Visits made at a user's request (ChatGPT-User, Perplexity-User, Claude-User) aren't scored either way. OpenAI says that for these visits "robots.txt rules may not apply", and Perplexity says the same of its user-requested fetches. What we saw: no rules for ChatGPT-User, Perplexity-User or Claude-User, and our test requests got the full page (fictional). If a firewall refuses them, we note it and suggest a settings change, but it doesn't affect the score.

What the main AI crawlers are for (recorded; only the search crawlers are scored):

Only GPTBot is blocked. That doesn't stop your site appearing in ChatGPT search, so it isn't marked down; it's a choice for you rather than a fix. The one scored issue is the noindex tag on the combined services page, which also explains why that page isn't in Google's index (section 10).

Our test checks robots.txt and requests sent under each crawler's name from our own computer, which uses a data-centre connection. Real crawlers come from their operators' published addresses, so a refusal we see can be a false alarm, and our test can't fully confirm a firewall that filters by address. If we find a possible block, we mark it "confirm with your host": the fix, if one is needed, is usually a settings change by whoever manages your hosting or security service, and we'd confirm it with them before changing anything. If you use a security service, ask it to allow the crawlers' published address ranges.

10. Other website basics (not scored)

Sample report: fictional business. All figures in this section are made up.

These checks don't add to the readiness score, but a problem here can hold back everything else.

None beyond the items above.

11. Prioritised fix list

Sample report: fictional business. All figures in this section are made up.

Ordered by impact first (High, then Medium, then Low), then by effort (Low, then Medium, then High). "weCDEF" means we'd make the change during a pilot if you give us access, and only after you approve it in writing. "You" means your office would make it, with our guidance; that includes anything you'd rather not give us access to.

12. Method and limits

Sample report: fictional business. All figures in this section are made up.

What we did

• Ran each prompt in prompt set WP (waterproofing) (version 1.0) 3 times on ChatGPT and 3 times on Perplexity, each in a fresh signed-out session, on the dates shown in section 2.
• For every run, logged the tool, the model name if the tool showed one, that we were signed out, the connection country, the date and time (Singapore time), and the exact prompt, and saved a screenshot of the full answer.
• Scored the readiness checklist in sections 3 to 9 using the weCDEF audit rubric (version v2), from public information only. We didn't log in to any of your accounts.
• Counted the outcome from the run log: the mention rate per question and tool, and share of voice against every other firm named in the answers.
• Used the same prompts, rubric and checklist we use for every firm in your trade.

Two separate measures

• Readiness score: weCDEF's own 100-point checklist (rubric v2, not an industry standard) of what your firm has in place that AI tools and search engines can find, across seven areas: search presence (10 points), reviews (20), structured data and FAQs (10), content clarity (20), third-party mentions (20), licences and registrations (10) and AI crawler access (10). It's what our fixes work on. The weights are our judgement, not a measured effect, and we'll review them after 3 months of results.
• Items we couldn't check: these are marked Not checked and left out, so that area is scored only on the items we could check, scaled to its full points. If a whole area couldn't be checked, the score is re-scaled over the other areas and marked provisional until we check it.
• Outcome: what ChatGPT and Perplexity actually said, measured as how many of 3 runs named you for each question, and your share of all firm mentions against each named competitor.

The outcome is not part of the readiness score. A better readiness score doesn't automatically mean a better outcome, which is why we measure both.

How the AI checks were run

The checks were run by a weCDEF team member, by hand, on a computer connecting to the internet from Singapore, while signed out, in a fresh session for every run, so the tools answer as they would for someone searching in Singapore. Every prompt also names Singapore. Each prompt was run 3 times on each tool. This sample covers ChatGPT and Perplexity.

Limits

• The readiness score is a structured checklist, not a precise measurement. It counts visible signals that, in our judgement, help search engines and AI tools find and confirm your details, based on what Google, OpenAI and Perplexity publish. It doesn't measure what any tool will actually say about you.
• AI answers vary. The same question can get a different answer for a different person, location, device, account or time of day, and tools change their models without notice. Running each prompt 3 times reduces this but doesn't remove it.
• Screenshots are a dated snapshot. They show what each tool said on the date and time recorded, not what it says today.
• Signed out, from a Singapore connection. Homeowners who are signed in may see different answers from ours, shaped by their own history and settings. Every prompt names Singapore, and every run is signed out, to keep the comparison fair over time.
• The search check is a stand-in. Section 3 uses one search tool's top 5 results, not a full Google results page, and rankings vary by device, location and personal settings.
• Review figures can lag. They come from the source named in section 4 and may be weeks behind Google Maps.
• Registers. Section 8 reports what each register showed on the date we checked. No AI company says it uses these registers, and nothing here is advice on which registrations you need.
• Crawler tests are indirect. Section 9's requests come from a data-centre connection, not from the real crawlers, so a possible block is confirmed with your host before anything is changed.
• Public information only. We can't see your Google Business Profile insights, website analytics or enquiry numbers unless you share them.
• Competitors are compared only on what's visible (how often they're named and their published rating). They are not fully scored.
• Small samples. 3 runs per question is enough to see patterns, not to rank firms precisely. Share of voice depends on which competitors the tools happen to name.
• Changes over time can come from our fixes, from changes the AI companies make, or both. We report what changed; we can't always say why.
• Nothing in this report is legal advice.

13. Next step: a free 60-day pilot

Sample report: fictional business. All figures in this section are made up.

If you'd like to see whether these fixes make a measurable difference, we're offering a small number of free 60-day pilots.

• What happens: we agree the fixes from section 11 with you. Where you give us access (your Google Business Profile, website pages, structured data, directory listings), we make the changes ourselves, after you approve each one in writing. Where you'd rather not give access, we tell you exactly what to change.
• How we measure: we re-run the same 4 prompts, 3 times each on the same tools, at about day 30 and day 60, with dated screenshots, re-score the readiness checklist, and send you a short report comparing the outcome and readiness score with this one.
• What it costs: nothing during the pilot. You give us the access you're comfortable with and a little of your time.
• After the pilot: you decide whether to continue. Ongoing work would be quoted to you in writing first, with the same re-check every month. There's no obligation.

To start, reply to this report or contact us at contact@wecdef.com.

14. No guarantee

Sample report: fictional business. All figures in this section are made up.

We can't promise that ChatGPT, Perplexity, Google AI Overviews or any other AI tool or search engine will name your firm, link to your website or rank you first, or that you'll get more traffic, enquiries or jobs. These tools are run by other companies, outside our control. They decide what to show, they change without notice, and their answers vary from person to person and day to day. This report shows what we saw on the dates and with the prompts recorded. What we do promise is to run the same prompts the same way each month, keep dated screenshots, and report what changed, whether it went up, down or stayed the same.

This free audit is general information based on public data, given without charge or obligation. It isn't a guarantee of any result (see above).

ChatGPT is a trademark of OpenAI. Perplexity is a trademark of Perplexity AI, Inc. Google is a trademark of Google LLC. weCDEF isn't affiliated with, endorsed by or sponsored by any of them.

Appendix A: Screenshot index

Sample report: fictional business. All figures in this section are made up.

End of sample report. Fictional business; no real results.