What Paywall CTA Copy Changes Have Real Companies Tested?
- By Ali Abouelatta
- n = 795
- Published
- Updated
- 2 min read
- paywall · monetization · experiments · mobile · ux-patterns · pricing
Short answer
Across 795 detected before/after paywall CTA changes from 146 companies tracked by Lazyweb Research, 344 (43%) changed the CTA element itself and 304 (38%) rewrote the actual button copy.source[1] The most common landing button after a rewrite is a plain "Continue" (34 of 304 rewrites, 28 companies), followed by "Subscribe" and "Start free trial."source[1] These are detected UI diffs with inferred rationale, not measured A/B results.
Of 795 detected paywall CTA changes across 146 companies, 304 (38%) rewrote the button copy itself — Lazyweb Research, July 2026.
Bar chart · Rows
Rows — Where the button copy lands after a rewrite
One block is about 1.2 Rows
- continue
- 34
- subscribe
- 17
- start free trial
- 16
- try for free
- 10
The finding
Lazyweb Research tracked 795 detected before/after paywall CTA changes across 146 companies.source[1] Of those, 344 (43%) flagged the CTA element itself as changed, and 304 (38%) had button copy that was actually rewritten (before and after text both present and different).source[1] Every one of the 795 rows carries an LLM-inferred rationale, so this is a census of what changed, not a leaderboard of what won.
Where the button copy lands after a rewrite
Among the 304 rewritten buttons, the most common resulting copy is short and generic:source[1]
| Button after rewrite | Rows | Companies |
|---|---|---|
| continue | 34 | 28 |
| subscribe | 17 | 11 |
| start free trial | 16 | 7 |
| try for free | 10 | 6 |
Even when apps rewrite the button, a large share land back on a neutral "Continue."
How to apply it
Treat CTA copy as one lever among several: only 344 of 795 changes (43%) touched the CTA element, so most paywall iterations move price, trials, plan order, or reassurance instead.source[1] If you are auditing your own paywall, benchmark the bundle of changes, not just the button word.
Caveats
All 795 rows are detected before/after diffs with inferred rationale, never measured lift.source[1] Some of the 304 "text differ" rows are annotation noise (e.g. a caption describing the button) rather than real copy swaps, so treat 304 as an upper bound on genuine rewrites.
Real in-market screens
Screens from real, in-market apps, tracked by Lazyweb Research.
Grammarly1125 × 2436opens the full image 
Headway1179 × 2556opens the full image 
Raycast1206 × 2622opens the full image 
Tinder1179 × 2556opens the full image
The numbers
| # | Stat | Computed from |
|---|---|---|
| [1] | 795 | Total detected before/after paywall CTA experiment rows (paywall_cta_experiments). |
| [2] | 146 | Distinct companies with at least one detected paywall CTA change. |
| [3] | 344 (43%) | Rows where cta_changed = true; 344/795. |
| [4] | 304 (38%) | Rows where cta_text_before and cta_text_after both present and differ; 304/795. |
| [5] | 34 rows / 28 companies | Most common cta_text_after among 304 rewrites: 'continue'. |
| [6] | 17 rows / 11 companies | 'subscribe' as cta_text_after among 304 rewrites. |
| [7] | 16 rows / 7 companies | 'start free trial' as cta_text_after among 304 rewrites. |
| [8] | 10 rows / 6 companies | 'try for free' as cta_text_after among 304 rewrites. |
Methodology
Universe: 795 detected before/after paywall CTA changes across 146 companies within the ~800-tracked-app mobile corpus, July 2026. Extraction: automated before/after screen diffing with LLM-inferred rationale. Key caveat: these are detected UI diffs, not measured A/B outcomes.
Sources
- [1]
Lazyweb Research analysis of 795 detected paywall CTA changes (146 companies, ~800-app mobile corpus), July 2026
Detected before/after UI diffs with LLM-inferred rationale (the 'learning' field), not measured A/B outcomes. Some cta_text rows contain descriptive annotations rather than pure button copy; theme tags are regex-over-LLM-text and approximate.
Source: Lazyweb Research — proprietary analysis of real, in-market app screens. Cite as Lazyweb Research, 2026-07-07.