Comparison

Straight Edge GPT vs ChatGPT

ChatGPT is a generalist optimised to produce a good answer. Straight Edge GPT is a specialist optimised to produce a checkable one. Here is exactly where that difference shows up — hallucination.

Feature by feature

CapabilityStraight Edge GPTChatGPT
Citations on factual claimsEvery claim carries an openable URLOnly when browsing triggers, often paraphrased
Confidence score0–100 with the basis stated in one lineNone — tone is equally confident either way
Says "I don't know"Required opening line below the confidence floorRare; usually produces a plausible answer instead
Source rankingFilings, papers and raw data outrank blog summariesWhatever the retriever surfaced, unranked
Contradiction detectionConflicting sources shown side by side with the discrepancyConflicts silently averaged into one smooth claim
Tool-use transparencyEvery query and page opened is shown in fullCollapsed or hidden
Exportable reasoning traceFull JSON download, auditable by a third partyNot available
Speculative padding"Some believe", "it could be" — banned unless requestedCommon hedging filler
Ecosystem breadthFocused on verified answeringImages, voice, apps, huge plugin ecosystem

Comparison reflects default behaviour on typical research prompts. ChatGPT behaviour varies by model and whether browsing is triggered.

Four hallucinations, and what happens here instead

> How many people does the WHO estimate die from air pollution each year?

Typical generalist

"Around 7 million people die each year from air pollution." Stated flatly, no scope, no link.

Straight Edge GPT

7 million combined ambient and household; ambient alone is 4.2 million. Contradiction logged: two outlets quote different figures for the same claim because they use different scopes. Confidence 94/100, linked to the WHO fact sheet.

> What did the court rule in [an obscure 2023 case]?

Typical generalist

Invents a plausible holding, a plausible judge and a plausible docket number. All three can be fabricated.

Straight Edge GPT

"I don't know the answer to that, but here's what I think…" — searches, finds no primary record, declares the gap, and refuses to name a holding it cannot open.

> Cite three studies showing X works.

Typical generalist

Returns three citations, one or two of which do not exist or don't say what's claimed.

Straight Edge GPT

Opens each paper before citing it. Anything it could not retrieve is dropped, and the answer says how many candidates failed verification.

> What's the current price / latest release / who won last night?

Typical generalist

Answers from a training cut-off unless browsing kicks in, with no warning that it's stale.

Straight Edge GPT

Anything checkable is checked live at answer time, and the retrieval timestamp is part of the verdict.

Test it on a question you already know the answer to

That's the honest benchmark. Three free answers, no card, no account. Then $19.99/mo, or $9.99/mo billed yearly.