- Home
- Skills
- Research & Analysis
- Unprompted: What AI Assistants Actually Recommend
Unprompted: What AI Assistants Actually Recommend
Real, weekly counts of which brands ChatGPT, Claude, Gemini and Perplexity name for buying questions. Free, no API key.
Free
Works with the AI Agents you already use
Unprompted: What AI Assistants Actually Recommend
Example session with this skill installed
Which AI image generators do AI assistants recommend most right now? Is Midjourney or DALL-E named more?
- Read your context and instructions
- Compiled the unprompted
Midjourney is named far more often than DALL-E. Across all six assistants combined, Midjourney appeared in 56% of answers (251 of 450) and DALL-E in 18% (80 of 450). Neither one is the most-recommended overall, though. Adobe Firefly leads.
AI Image Generators · measured week of 2026-09-28 · 450 answers from chatgpt, claude, claude-code, codex, gemini, perplexity
- Adobe Firefly: named first in 24% (109 of 450), named at all in 57% (257 of 450)
- Imagen: named first in 16% (72 of 450), named at all in 27% (123 of 450)
- Midjourney: named first in 14% (63 of 450), named at all in 56% (251 of 450)
- Canva: named first in 10% (47 of 450), named at all in 28% (126 of 450)
- Flux: named first in 10% (43 of 450), named at all in 34% (153 of 450)
- DALL-E: named first in 8% (34 of 450), named at all in 18% (80 of 450)
Midjourney vs DALL-E: Midjourney comes out ahead on both measures. On "named first" it leads 63 to 34. On "named at all" it appears about three times as often.
Mentions vs first place: Midjourney is named nearly as often as Firefly (251 vs 257 answers), but it's named first much less often (63 vs 109). Imagen is the opposite: it appears less often overall but is often the first name given. Ideogram shows up in 42% of answers (187 of 450) but is first in only 8%.
- "Named first" means first among the brands the chart tracks. It isn't always the first name in the answer.
- These are counts of what the assistants said, not quality ratings. The figures combine all six assistants, so they don't show any single assistant's habits.
- The standings label the data as week 2026-09-28 but were read on 2026-10-28, so the newest numbers are a few weeks old.
Source: Unprompted, https://unprompted.report/chart/ai-image-generators
I can also break this down for one assistant, for example what ChatGPT on its own names.
Connects securely to your tools. The creator never sees your data.
About this skill
Ask your agent "which AI coding assistant does ChatGPT recommend most?" and it will usually guess, or run one prompt once and call it a trend. This skill gives it real numbers instead.
It reads Unprompted, a free public chart that asks AI assistants the same real buying questions five times each, every week, and counts which brands they name. Your agent gets the share of answers that named each brand, the share that named it first, the sample size and the week measured, and quotes them correctly.
What it does
- Answers "which brands do AI assistants recommend for X?" with counted, dated figures
- Compares two brands on how often assistants name them, and how often first
- Breaks results down for one assistant (just ChatGPT, say) from the published raw answers
- Links straight to a brand's page and a side by side comparison
- Says plainly when a category is not measured or a brand is not on the chart, instead of making numbers up
- Never calls a brand "best": these are counts of what assistants said, not ratings
Categories today: AI coding assistants, AI writing tools, AI image generators. New categories show up automatically, because the skill reads the live category list each time.
Who it is for: marketers, founders and SEO teams working on AI visibility (GEO / AEO) who want a neutral, repeatable benchmark rather than a single screenshot of one chat.
Requirements: an agent that can fetch a URL (web fetch or curl). Per assistant breakdowns also need it to run a small script (Python, Node or jq). No Unprompted API key or account needed. The skill and the data are free; your agent's usual costs still apply.
About the data: Unprompted is run by Skald Studio. No placement is for sale, and the method, every question and every raw answer are public at unprompted.report/methodology and on GitHub.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Research & Analysis.
Covers only the categories Unprompted measures (AI coding assistants, AI writing tools and AI image generators today). Figures are as fresh as the latest published week, which can be several weeks old. Shares count what assistants said, not product quality. Brands the chart does not track are not counted. Per assistant breakdowns need the agent to run a small script, or it falls back to the brand page.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 9 days ago
- Free with an Agensi account
- Software
- curlorweb fetch tool
- Optional
Python
orNode.js
orjq
Python, Node.js or jq gives per-assistant breakdowns.
Permissions
Network access is used only for read-only GET requests to unprompted.report and raw.githubusercontent.com (public data, no login). Terminal is used for curl and, optionally, a short script to count one assistant's answers. Nothing is written outside the working folder and no credentials are read.