Quarterly · Q2 2026

ProductsVerdict AI Report

We analyzed 100 AI tools across writing, video, code, agents, voice, and image generation. Here are the winners, the rising contenders, and the ones quietly dying — independent, no vendor briefings, no paid placements.

100 tools researched6 categories30-day uptime probesPublished 2026-06-01

Key findings

Writing winner: ChatGPT held its lead — barely

Across 18 standardized prompts (brief generation, long-form, rewriting, SEO outlines), ChatGPT scored 8.9/10 on output quality. Claude 3.7 closed to 8.7. Differentiator: tone-matching on long-form. Loser: Jasper, now trailing both on raw quality at a 4× price.

Agents went from novelty to default in 90 days

Q1 to Q2: 41% of tools we tracked shipped some form of agentic workflow (multi-step task execution). Only 12% deliver reliably beyond demos. Best-in-class: bespoke vertical agents. Worst: 'general' agent platforms that wrap GPT in scaffolding.

Pricing reset: per-seat is collapsing

Usage-based pricing grew from 22% → 47% of new launches this quarter. Per-seat SaaS pricing on AI tools is increasingly seen as misaligned. Vendors still on flat per-seat are losing mid-market deals to usage-based competitors.

The quiet die-off: 14 tools we recommended last year shut down

Of 87 AI products in our 2025 best-of lists, 14 (16%) have shut down, been acquired and gutted, or pivoted away from their original use case. We've removed them from rankings and added the survivors-only filter to comparisons.

Category leaderboard

One winner, one value pick, one rising contender, and one tool losing ground per category.

Category	Best overall	Best value	Rising	Losing ground
AI Writing	ChatGPT	Claude	Perplexity Pages	Jasper
AI Video	Synthesia	HeyGen	Veo 3	Pictory
AI Code	Cursor	Cline	Devin (V2)	Replit Ghostwriter
AI Agents	Lindy	n8n + GPT	Stack AI	AutoGPT forks
AI Voice	ElevenLabs	PlayHT	Cartesia	Murf legacy
AI Image	Midjourney	Flux Pro	Recraft	older DALL-E flows

Methodology

•100 AI tools tracked across 6 categories — selected by Q1 traffic, capital raised, and reader requests.
•Standardized prompt suite per category (18 prompts for writing, 12 for video, 15 for code, etc.). Same prompts, scored blind.
•Pricing pulled from public pages on 2026-05-15. Per-seat costs normalized to a 5-seat team.
•Independent uptime probes across 30 days for hosted tools.
•User sentiment cross-checked across G2, Reddit, and our own moderated review pool — fake/incentivized reviews filtered out.

Full raw data and per-prompt scoring are available on request to verified researchers and journalists. Email research@productsverdict.com.

Get the next report first

The Q3 2026 ProductsVerdict AI Report drops in September. Subscribers get it 7 days early.

Subscribe to the newsletter

Previous reports

· Q1 2026 — The AI Agent Reality Check (archived)
· Q4 2025 — Year-in-Review: AI Tools That Actually Made Money (archived)