SingaporeAssignmentHelp Expert Profiles: AI Audit

Short answer: A 15 July 2026 audit of all seven populated expert-directory pages found 77 public profiles, despite the directory's broader 2,100+ writer claim. Twenty-three profile associations fall into 11 groups using byte-identical portrait files. One live portrait contains trusted OpenAI/ChatGPT C2PA provenance identifying GPT-4o and AI-created media, although its current outer wrapper has validation warnings. The pages also contain repeated testimonial blocks and internal biography inconsistencies. These findings justify asking how profiles are verified; they do not prove that every expert is fictional or unqualified.

Commercial disclosure and method — 15 July 2026: Codingo competes with SingaporeAssignmentHelp.com and commissioned this exact-domain audit. We crawled the public directory and profile pages, compared image-file hashes, inspected C2PA Content Credentials with c2patool and checked selected programme claims against institution pages. We did not contact the named people or universities, access private records, place an order or test delivered work.

Findings at a Glance

1. How Many Experts Are Publicly Verifiable?

The experts directory advertises 2,100+ Singapore-based writers and more than 81,080 orders. We enumerated every populated pagination page and found 77 public profiles.

That is not automatically a contradiction. A marketplace may keep much of its roster private or inactive. It does mean a buyer cannot use the public directory as verification of the complete 2,100+ claim. Ask how many experts are currently active, how many are based in Singapore, and what evidence supports those counts.

The 77 displayed per-profile order counters add to 155,096. That exceeds the directory's 81,080+ headline. The site may be counting different periods, categories or repeat allocations, but it does not explain the definitions. A useful correction would label the denominator and time period for each metric.

2. Eleven Portrait Files Are Reused Across Different Names

We downloaded the live portrait assets and compared SHA-256 hashes. Eleven identical-file groups cover 23 profile associations:

Different experts can legitimately use stock, illustrative or privacy-protecting images. The problem is disclosure: these cards label the images generically as expert images and do not state that portraits may be generated, illustrative or shared. A customer could reasonably read each headshot as representing that named expert.

3. One Live Portrait Contains GPT-4o Content Credentials

The Baozhai Tan profile loads a live portrait file whose URL filename is “Roisin-Whelan.png”. The file downloaded on 15 July 2026 had SHA-256:

6937fbab7980b1adf0e3e5bb47ab59980ec443b2f4632dfbc429e2a8618f8623

Its embedded C2PA ingredient manifest identifies:

• claim generator: ChatGPT;
• creation software: GPT-4o;
• digital source type: trained algorithmic media;
• conversion software: OpenAI API; and
• issuer: OpenAI, with a Truepic Lens CLI in ChatGPT common name.

Using c2patool v0.26.60 with the official C2PA trust list, the validator reported the OpenAI signer as trusted and validated the ingredient signature, hashed references, asset data hash and ingredient manifest. The overall current wrapper still reports invalid because its signing credential has expired and its first action is malformed. The narrow, supportable conclusion is: the portrait contains trusted OpenAI/ChatGPT provenance identifying GPT-4o and AI-created media, while the present outer wrapper is not fully valid.

OpenAI says GPT-4o-generated images carry C2PA metadata identifying GPT-4o. The C2PA specification defines “trainedAlgorithmicMedia” for media created by a trained AI or machine-learning model. Content Credentials describe a file's provenance; they do not establish whether a person exists or whether a biography is true.

4. A Second Generated-Image Signal Is Reused Three Times

The same byte-identical portrait appears on the David Thomson, Yannik Cheang and Raphael Liao profiles. Its embedded manifest names Black Forest Labs, Flux.1 and AI-generated content. The image hashes and a trusted timestamp validate, but c2patool reports a claim-signature mismatch. We therefore treat this as strong provenance signals, not fully authenticated cryptographic proof.

The more basic fact does not depend on provenance metadata: all three public identities use the same image bytes. The provider could eliminate ambiguity by marking such images “illustrative” and offering a separate, privacy-safe credential-verification method.

5. Biography and Qualification Details Do Not Always Reconcile

Several profile pages contain internal or current-programme mismatches:

• Gordon Teo's page associates an MS in Statistics with James Cook University in one area and Ball State University in another.
• Javier Peng's headline material describes a master's qualification in Ancient History, while the biography refers to a PhD.
• Kevin Liong's Public Relations profile contains biography text about ancient history.
• Raphael Liao's media and communications profile includes a line about expertise for medicine students.
• Yannik Cheang's profile attributes a healthcare qualification to Singapore Bible College, while the college's published programme areas concern theology, biblical studies, counselling and music.
• Nicholas Zeng's profile refers to a master's in medicine from Duke-NUS, whose principal medical qualification is the Doctor of Medicine.

Programme names change, biographies can contain editing mistakes and older awards may not appear on current course pages. These are verification prompts, not findings that a credential is false. The provider should correct copy errors and state the exact award, institution, year and verification method.

6. Testimonials and Ratings Look Templated

The first 12 individual profiles checked used the same testimonial section. The expert directory displayed 24 testimonials, all dated Saturday, 4 May 2024. Sixty-six of 77 cards showed exactly 5/5; the remaining 11 showed exactly 4/5.

A shared site-wide testimonial carousel is possible. If so, it should be labelled as general service feedback rather than evidence about the named expert. The reviewed profiles did not show an expert-specific sample size, methodology or link to an external review record.

7. “Hire Directly” Versus the Agency Terms

Directory calls to action invite users to hire an expert. The terms say the Agency allocates a “Principal” and prohibit direct customer contact with that Principal. This distinction matters: selecting a public profile is not necessarily the same as contracting with or speaking directly to that person.

Ask whether the named expert is guaranteed to receive the order, whether direct questions are allowed, and what happens if the allocated specialist differs from the profile selected.

What This Audit Does Not Prove

• It does not prove that any named person is fictitious.
• It does not prove that a qualification is forged.
• It does not prove that every portrait is AI-generated.
• It does not prove that the claimed private roster does not exist.
• It does not test the quality of any completed work.

It does show that the public profiles are not adequate on their own to verify identity, credentials, portrait authenticity or the advertised roster size.

Questions to Ask Before Selecting an Expert

1. Is the portrait a photograph of the named person, a generated image or an illustration?
2. What exact qualification, institution, graduation year and subject specialism have been verified?
3. Will this named expert personally receive the order?
4. Can the provider verify credentials privately without exposing sensitive documents?
5. Are the testimonials specific to this expert, and what is their sample size?
6. May the customer ask the assigned expert technical questions directly?
7. What happens if the expert's skills do not match the brief?

FAQ

Are SingaporeAssignmentHelp.com's expert photos AI-generated?

At least one live portrait contains trusted OpenAI/ChatGPT C2PA provenance identifying GPT-4o and AI-created media. A second reused image contains strong Flux.1 signals but has a signature mismatch. This audit does not label every portrait AI-generated.

Does an AI portrait mean the expert is fake?

No. A real person may use a generated or illustrative image for privacy. The concern is that the profile does not disclose that distinction or provide another visible identity-verification route.

Why do duplicate portraits matter?

Twenty-three profile associations share images across different names. Without an “illustrative image” label, the cards can create a misleading impression that each portrait depicts its named expert.

Can SingaporeAssignmentHelp.com correct this audit?

Yes. Codingo will evaluate dated source corrections, explanations of image policy and public verification material, then record any material update.

Related Investigations

• Why Jaipur and Suntec both appear
• Checking the 15,840-review claim
• Guarantees compared with the site's own terms
• Privacy claims compared with forms and terms
• Five-minute provider verification checklist

Sources Checked

• SingaporeAssignmentHelp.com expert directory and the individual profiles linked above
• SingaporeAssignmentHelp.com terms
• OpenAI: Introducing 4o Image Generation and C2PA in OpenAI-generated images
• C2PA actions specification and trust-list guidance
• Singapore Bible College and Duke-NUS MD programme