Case studies

Is Raycast visible in AI answers? We asked 11 buyer questions, it was missing from 2

SeenByLLMs · 2026-10-16 · An outside test of public AI answers. We are not affiliated with Raycast.

When buyers ask an AI model for the best macOS launchers and productivity apps, does it name Raycast? We put 11 buyer questions to one AI model in a single run on 16 October 2026. Raycast was absent from 2 of 11 answers, scored 67/100, and took 28% share of voice against Alfred's 25%. It is on the AI shortlist, but it does not own it.

How we tested

We wrote 11 questions the way a buyer would type them, covering the macOS launcher and productivity category: alternatives queries, best-of queries, comparison queries and niche job-to-be-done queries. We tracked four named competitors alongside Raycast: Alfred, Spotlight, LaunchBar and Quicksilver. Every question went to Nemotron 3.5 Lightning (free) through OpenRouter with OpenRouter web search turned on. One model, one run, one date: 16 October 2026. A brand counts as "named" when its name or a known alias appears as a plain text match in the answer. Rank is the order of first appearance in that answer, not prominence in the layout. A page counts as a pitch target when an answer cites it and it names a rival but not Raycast. We excluded competitor domains, raycast.com, and platforms like GitHub, Reddit, YouTube and Wikipedia from that list.

The limits are worth stating plainly. This is one free model, one run, 11 questions. Read it as a signal about where AI answers point buyers, not as a ranking of the category. No answer came back empty, so every gap below is a real answer that chose someone else.

Results, question by question

Raycast appeared in nine answers and ranked first in five of them.

The pattern is clear. Raycast wins specific, job-shaped questions: extensions, clipboard plus launcher, free options, power users. It loses, or slips, on the generic head terms. On "best macOS launchers" it came fourth, behind Spotlight, Alfred and LaunchBar, even though it was named. And it disappeared entirely on two questions where the model did not think in terms of launchers at all.

Where Raycast went missing

"Which tools help Mac users run snippets and workflows quickly?"

The answer named Alfred (Powerpack) first, then CommandIt and Snippety. Only Alfred came from our tracked set, so this was not a sweep by rivals. The model read "snippets and workflows" as a text-expansion job and reached for dedicated snippet tools. Raycast's snippet and workflow features never entered the frame, which suggests the model does not associate the brand with that job yet.

"Which Mac apps do indie developers swear by for speed?"

The answer listed OpenMark, Brow, SQLAgent, iA Writer and Paper. No launcher appeared at all, from any tracked competitor. This is not a competitor win. The question asks about a personal toolset, and the model answered with writing and agent tools instead of the launcher category. One cited page was a dev.to post about making a Mac launcher feel instant, yet the answer still named no launcher. Raycast was not beaten here; the category was never in the running.

The pages behind those answers

Two cited pages named a rival but not Raycast. Both come from the snippets question, and both are the snippet tools' own product sites:

• commandit.ai - CommandIt | A keyboard-first command palette for developers, built MCP-native
• snippety.app - Snippety - Snippets Manager & Text Expander - macOS & iOS

Neither will list Raycast. The useful source here is the one already working for Raycast: brow-app.com was the most-cited domain in the run (6 citations) and already names Raycast. Editorial coverage came from xda-developers.com (3 citations) and theverge.com (2), including The Verge's Raycast how-to.

Why this matters

These are buying-intent questions. "Best X for Y" and "alternatives to Z" are what someone types shortly before they install something. When the model answers without Raycast, the buyer does not get a link to raycast.com and never arrives. Nothing in a normal analytics stack records that loss, because there is no session to attribute. Look at the two gaps above: one buyer was told to compare CommandIt and Snippety for snippets, another was handed a developer toolset with no launcher in it. Those are install decisions made without Raycast in the room, and they leave no trace on the site.

What Raycast could do next

1. Snippets gap. Publish a page titled around "snippets and workflows on Mac" that shows Raycast's snippet and text-expansion features directly. The answer was built from dedicated snippet-tool pages and Alfred's Powerpack pages (alfredapp.com was cited three times in that answer).
2. Snippets gap, comparison. Pitch xda-developers.com, which already covers Mac launchers and Alfred, for a snippet or text-expander roundup that includes Raycast.
3. Protect the strongest source. brow-app.com is the most-cited domain (6 citations) and names Raycast. Ask for an updated comparison or higher placement.
4. Indie-developer gap. This answer cited developer blog posts about native Mac apps (openmarkapp.com, brow-app.com, sqlagent.app, paper.pro, dev.to). A Raycast post in that format, about speed for indie developers, is the page type the model read.

Run the same check on your brand

1. Create an OpenRouter key at openrouter.ai/keys and set a spend limit, then paste it into the app. The key stays in your browser and is never sent to a server.
2. Enter your brand, up to five competitors and the questions your buyers actually ask. If you do not have a list yet, start from the suggested prompts in the app.
3. Press Run, then read which questions you lose, who was named instead, and which cited pages are worth pitching. Export the result as HTML or CSV when you are done.

This run cost about $0.08 on our own OpenRouter key and took 0.7 minutes.

Run a free check