Teardowns › Pricing & paywalls › Reference price

ChatGPT puts an $8 step between Free and $20 Plus

ChatGPT · Pricing page · Logged-out visitors (USD shown) · chatgpt.com · captured 16 Oct 2026

Honest nudgeReference price Zero-price effect

ChatGPT Pricing page, captured 16 Oct 2026 — Our capture, 16 Oct 2026, logged out. Numbers match the steps.

Step by step

1
Zero-price effect
Free is "$0 / month" and leads with "Unlimited text chats".
2
Reference price
Go is "$8 / month" with "Everything in Free and:" more messages, uploads and memory.
3
Friction
The Go card discloses "This plan may include ads."
4
Reference price
Plus follows at "$20 / month" with "Everything in Go and:".

Prices quoted in dollars ($), exactly as shown on the page we captured (US-priced). Prices vary by region and change over time.

The psychology

Why it nudges

A middle $8 tier likely changes the comparison: from Go, Plus reads as a $12 step rather than a jump from $0 to $20 (the neighbouring tier acting as a reference price). That reading is ours. The "Everything in… and:" chains make each plan read as an extension of the last. The tier taglines rotated between our captures ("Best for…" vs "Expanded access"). Whether this changes plan mix for OpenAI is not public.

● Mixed evidence (as used here) · More on reference price, with sources →

Context

An AI assistant with a large free user base and several paid tiers.

Honest nudge or dark pattern?

Honest nudge

Honest nudge: the trade-offs are disclosed on the card.

Steal this, honestly

Make each tier "everything in the last one, plus…", so the steps are easy to read.
Disclose trade-offs (like ads) on the card itself, as this page does.

What to test: Three tiers vs two (no middle tier). Measure the share choosing Plus.

Next: run it on your product

Apply this with your agent

Take the steps above into your own flow. The prompt carries this teardown, the honest-vs-dark line and the steps.

You are helping me apply a behavioural-design pattern ethically.

Pattern: ChatGPT: ChatGPT puts an $8 step between Free and $20 Plus
Bias: Reference price: A price is judged against a comparison price: a 'was' price, a monthly rate, a renewal rate.
Why it nudges: A middle $8 tier likely changes the comparison: from Go, Plus reads as a $12 step rather than a jump from $0 to $20 (the neighbouring tier acting as a reference price). That reading is ours. The "Everything in… and:" chains make each plan read as an extension of the last. The tier taglines rotated between our captures ("Best for…" vs "Expanded access"). Whether this changes plan mix for OpenAI is not public.
Our verdict: Honest nudge. Honest nudge: the trade-offs are disclosed on the card.
Source page: https://nudgeswipe.com/examples/chatgpt-pricing-ladder-free-go-plus/ (JSON: https://nudgeswipe.com/api/teardowns/chatgpt-pricing-ladder-free-go-plus.json)

Rules: say a brand "uses" a tactic, never that it "worked"; outcomes are not public unless the company published them; no invented uplift estimates; flag mixed evidence. Honest nudge = true, visible, easy to reverse, and in the user's interest. Dark pattern = hides costs, fakes scarcity or popularity, shames, or makes leaving harder than joining. Quote company-published results as "X reports", with the source link, and don’t generalise them. Don’t cite retracted or failed-to-replicate studies as support (e.g. Ariely & Wertenbroch 2002). Don’t state study figures unless they’re on this page or in a source you can link. If an idea crosses the line, link the matching deceptive.design pattern.

Tasks:
1. Ask me for my product, the flow (URL or screenshots) and the goal.
2. Audit where this pattern already appears, or could, in that flow.
3. Propose 2-3 ethical experiments: hypothesis, variant copy or UI, primary metric, and a guardrail metric (refunds, complaints, cancellations).
4. Flag every dark-pattern risk and say how to stay on the honest side.

Open in ChatGPT Open in Claude Open in Grok Open in Cursor

ChatGPT, Claude, Grok and Cursor open with the prompt filled in (it is also copied, in case the prefill doesn't load). Gemini and Codex don't support prefilled links, so the prompt is copied and the app opens; paste it in.

Sources

Our capture of chatgpt.com, 16 Oct 2026
Kalyanaram & Winer (1995), Empirical generalizations from reference price research. doi.org/10.1287/mksc.14.3.G161

Related teardowns

economist.com

The Economist Pricing screenshot, captured 16 Oct 2026

Price & attribute framingHonest nudge

The Economist prices a year by the week, and preselects it

The Economist · Pricing · $ as captured

slack.com

Slack Pricing screenshot, captured 16 Oct 2026

Reference priceConditional

Slack shows the full price struck through next to the intro price

Slack · Pricing · $ as captured

spotify.com

Spotify Pricing screenshot, captured 16 Oct 2026

Zero-price effectConditional

Spotify leads with "£0" and spells out the price after

Spotify · Pricing · £ as captured

5 new nudges, every Friday

Five real product and marketing moments, the psychology behind each, and one thing to try. Weekly issues start soon. Your email is stored with our email provider (Resend) and used only for this newsletter.