AI tools: hype vs traction 2026

We index roughly one million tech articles a year to build our newsletters, which gives us a clean read on where press attention goes. The chart shows, for each AI tool, how many headlines out of every 10,000 we indexed named it between July 2025 and August 2026. ChatGPT still leads, but the story is the trajectory underneath it. ChatGPT is fading, from 133.9 mentions per 10,000 in the first half of the window to 91.9 in the second. Claude is surging, from 29.3 to 114.4, and in the two most recent complete months, July and August 2026, Claude was the single most-named tool at 124.2 per 10,000, ahead of ChatGPT at 99.3 and Gemini at 78.2.

press mentions per 10,000 tech-news headlines, July 2025 to August 2026

Share of coverage over the mature window. Below the top three, coverage thins fast: Grok, Copilot, and Perplexity together get less than half of ChatGPT's share, and DeepSeek (16.7) is rising while Perplexity is falling.

Source: Dupple news radar, September 17, 2026. Method: case-insensitive substring match of each tool name against tech-news headlines, divided by all headlines indexed in the same window. We report share of coverage rather than raw counts, and limit the window to July 2025 onward, when our index reached full volume. Common-word names such as Runway were excluded from the ranking because they collide with unrelated coverage.

This is the study. For each tool we line up all three of our datasets side by side: press share of coverage, the number of emerging open-source projects in our tracker built around it, and its Toolradar user rating with review count. Read across any row and the tool's standing changes with the column. ChatGPT wins press and ties for the top rating, but trails Claude four to one on developer projects. Claude wins developers and is climbing the press fast, but its blended rating is dragged to 3.0. ElevenLabs is barely covered and has no open-source footprint, yet it is one of the highest-rated tools on the page.

Source: Dupple news radar, Dupple GitHub trending tracker, and Toolradar catalog, September 17, 2026. Press share covers July 2025 to August 2026. Emerging OSS projects = tracked repositories in the 20 to 1,000 star range whose name, topics, or description contain the tool name; "closed source" marks tools with no distinct open-source project name to count, and 0 marks tools we verified have no tracked open-source projects. Ratings are the Toolradar aggregate of external G2, Capterra, and Trustpilot scores for the tool's main catalog entry.

Our tracker follows repositories as they move through the 20 to 1,000 star range, which catches new tools while they are still emerging, before they graduate to fame. The flagship model repositories are long past that range, so what this measures is the ecosystem forming around each model: the agents, the connectors, the skills. Read that way, one model dominates. Of the 792 projects we currently track, 66 reference Claude, against 24 for OpenAI, 20 for Gemini, and 15 for ChatGPT. The fastest climbers are almost all agent frameworks and Model Context Protocol servers, the plumbing of the agentic layer.

• Claude is referenced in 66 of 792 tracked projects, more than OpenAI (24), Gemini (20), and ChatGPT (15) combined would suggest for a tool the press ranks third.
• The fastest-growing repository in the entire tracker over the last 30 days, up 714 stars, is a Claude-based autonomous-agent framework, and the only live breakout tied to a named model is a Claude and MCP server that added 129 stars in a single week.
• Midjourney, ElevenLabs, Replit, and Stable Diffusion are each referenced by zero tracked projects. Image, audio, voice, and closed coding tools have almost no open-source momentum, even where they are heavily used.
• The median tracked project gained just 1 star in 30 days, and only 5 repositories were flagged as breakouts in the last month, so the concentration around a handful of agent and MCP projects is the real signal.

Source: Dupple GitHub trending tracker, September 17, 2026. Method: repositories the tracker follows in the 20 to 1,000 star range, counted by whether their name, GitHub topics, or description contain the model name (case-insensitive); star growth is measured between the tracker's own snapshots. Ecosystem counts reflect projects built around a model, not the model's own repository.

The third dataset holds a surprise of its own. For the most-covered tools, business reviewers and consumer reviewers do not just disagree, they land three full points apart on a five-point scale. On G2, where reviews come from verified business software users, these tools score 4.2 to 4.8. On Trustpilot, an open consumer review site, the same brands score 1.4 to 1.8, usually over billing, refunds, and support. Toolradar blends the sources, which is why household names show middling aggregate scores. The tools that avoid the split, and sit at the top on both volume and rating, are the ones sold mostly to professionals.

• The two highest-rated tools with real volume, ChatGPT (4.6 from 2,958 reviews) and ElevenLabs (4.6 from 1,236), are rated almost entirely by business users, which is why they escape the consumer drag.
• Where consumers dominate the sample, the score collapses: Suno sits at 2.2 across 810 reviews because 802 of them are Trustpilot reviews averaging 1.6.

Source: Dupple analysis of Toolradar external review data, September 17, 2026. Method: per-source ratings and review counts as stored on each tool's main Toolradar catalog entry; the blended figure is Toolradar's own aggregate across all listed sources. "Not listed" means no Trustpilot profile was attached to that entry.