ChatGPT vs AlphaLens: VVV on September 17, 2026
"Just ask ChatGPT to research a stock" sounds reasonable until you need the answer to point at something. This is a dated, sourced look at what that gap actually looks like on one real name — Valvoline, a mid-cap that a systematic accrual screen flagged for a real, explainable divergence between earnings and cash flow.
Method
The test question, the kind an investor would actually type into a general AI chat: "Research this stock: Valvoline (VVV). What's the financial picture, key risks, and is it worth buying?" That is a fair proxy for how most people actually use chat tools for stock research — one open-ended prompt, no follow-up.
The AlphaLens side of this comparison is not a fresh run staged for this page — it's the same Earnings Quality Analyzer output already published and dated on the site: the Valvoline (VVV) case study, run September 2, 2026, built from SEC 10-K/10-Q filings, live market data, and current news as of that date. Anyone can open that page and check the figures below against it.
What each approach actually produces
General AI chat
Asked to research VVV cold, a general assistant with no standing research process behind it answers from whatever it has learned or can find — a fluent narrative covering the business, some risks, and a qualitative read on valuation.
What it has no built-in mechanism to do: run a systematic check for divergence between reported net income and operating cash flow, catch that Valvoline's FY2023 numbers show exactly that divergence, or explain it without being told to look for it. If the prompt doesn't mention the 2023 Global Products sale to Aramco by name, there's no guarantee the reply surfaces it — and no dated citation tying any number in the reply to a specific filing.
AlphaLens — Framework 04, Sept 2, 2026
Excerpt, quoted directly from the published run:
"The screen compared a fiscal year in which Valvoline reported roughly $1.4 billion of net income against operating cash flow that was slightly negative for the same period... It is the March 2023 sale of the Global Products segment to Aramco: a ~$2.65 billion transaction that booked a large one-time gain in net income."
Full framework output, including the "what to check next" section: norrisai.us/analysis/vvv-2026/#f4
What this does not prove
This is one paraphrased comparison against one dated framework run — not a controlled test across every chat model, account tier, or browsing setting. A general assistant with live web browsing turned on, prompted well, and cross-checked by hand can get closer to this. What this does show: AlphaLens' Framework 04 output for VVV is fully sourced to a named filing and a stated screen threshold, verifiable by anyone who opens the run page — and that property (check against live filings, then state the verdict, with nothing invented if the number isn't there) is what the framework is built to do on every ticker, not just this one.
Run the Earnings Quality Analyzer on any stock
Live SEC filings, live prices, current news — the same accrual and cash-conversion checks applied to VVV above, run on whatever ticker you're actually looking at.
Run Framework 04 on VVV →Related: Full VVV case study (all 15 frameworks) · Earnings Quality Analyzer · AlphaLens vs ChatGPT, Claude & Perplexity · Compare index