AgentsAutonomy & tool use
WSJ Columnist Describes Claude Code TV Watchlist as Survey Finds 41% of U.S. Viewers Use AI to Find Shows
A Wall Street Journal column describes using an AI coding tool, identified in its URL as Claude Code, to filter TV picks by rating, episode count and unwanted subjects, as Hub research finds 41% of U.S. viewers already use AI to find shows.
From the WSJ · Technology article of November 15, 2026 and cited sources: Hub Entertainment Research's October survey of 1,601 U.S. consumers found that 41% use general AI tools such as ChatGPT, Claude or Google AI to find shows. [2]
SOURCES[2] RBR: Hub — As AI Enters Search Box, Viewer Struggle Persists rbr.com
AI-assisted, not human-edited. Methodology Corrections Who owns this site
- What happened
- Hub Entertainment Research's October survey of 1,601 U.S. consumers found that 41% use general AI tools such as ChatGPT, Claude or Google AI to find shows. [2]
- Original source
- WSJ · Technology article, independent perspective
- Brief by
- AT Pulse research desk AI-assisted; publisher and contact on the desk page
- When
- November 15, 2026 article published
- Covered
- Nov 15 edition
- Evidence
- 7 sources across 7 sites
- 1How AI Helps Me Find the Right TV Shows at the Right Timeswsj.com
- 2RBR: Hub — As AI Enters Search Box, Viewer Struggle Persistsrbr.com
- 3To Vima (WSJ syndication): How AI Can Find the Perfect…tovima.com
- 4What's on Netflix: Netflix unveils TV interface overhaul…whats-on-netflix.com
- 5eWeek: Not Sure What to Watch Next? Just Ask Netflix's New…eweek.com
- 6Scientific American: How Claude Code Is Bringing Vibe…scientificamerican.com
Reach: not measured yet
Hub Entertainment Research's October survey of 1,601 U.S. consumers found that 41% use general AI tools such as ChatGPT, Claude or Google AI to find shows. Among 16–34-year-olds the figure is 54%. Of those who use AI this way, 57% say it works better than search engines. These are self-reported preferences, not measures of how accurate the recommendations are. [2]
The column filters on high ratings, many episodes and no subjects the writer dislikes. Those are structured criteria that match what viewers say they want: 30% call episode or season count very important before starting a series, rising to 40% among younger viewers. [1] [2]
A coding agent changes how the long-standing advice to prompt a chatbot for media picks gets carried out. Claude Code supports persistent instructions, connectors to outside data and scheduled tasks. In principle, those features could turn a one-off chat request into a reusable personal tool that refreshes itself. Whether the columnist used any of them is unknown. [3] [6] [7]
Streamers are building their own conversational discovery. Netflix's OpenAI-backed search lets users ask for content by mood, such as scary but not too scary. Meanwhile, recommendations from people and trailers still beat streaming algorithms for discovery, and 47% of viewers struggle to tell services' content apart. [2] [4] [5]
Hub's survey finds general AI assistants are already a common way to discover shows, especially among younger viewers.
Read the full assessment
The likely but unproven implication: streamers could lose control of recommendations, and clean data on ratings, episode counts and content warnings could become more valuable.
Executive brief
A Wall Street Journal column published Nov. 15, 2026 describes using an AI tool, which the article's URL identifies as Claude Code, to filter TV picks. The filters are high ratings, many episodes, and no subjects the writer dislikes (WSJ). The more consequential fact comes from elsewhere. Three days earlier, Hub Entertainment Research reported that 41% of U.S. viewers already use general AI tools to find shows, and 57% of those users say the tools work better than search engines (RBR/Hub). Only a feed excerpt of the column was reviewed, so how the tool was built is not known.
What changed and event timeline
WSJ sets the template
Alexandra Samuel's column showed how to paste viewing history from Netflix or Goodreads into ChatGPT or Claude.ai to get mood-based and household-aware picks ().
Netflix brings in OpenAI
Netflix announced conversational search built on OpenAI models, with its own catalog and user context layered on top (). It first appeared as an opt-in iOS beta ().
Coding agents reach non-programmers
Scientific American described non-coders using Claude Code to build working software from plain-English instructions ().
Survey measures adoption
Hub's October survey of 1,601 U.S. consumers found 41% use ChatGPT, Claude or Google AI to find shows, rising to 54% among 16–34-year-olds ().
WSJ watchlist column
The column describes a personal setup for finding highly rated, long-running series without unwanted subjects. The URL slug links it to Claude Code ().
Capabilities and access
- Tool: Claude Code, according to the article URL. The model and version used are not stated in the reviewed excerpt.
- What Claude Code does: It is Anthropic's agentic tool. It is available in the terminal, IDEs, a desktop app and the web (Claude Code docs).
- Access: Most surfaces need a Claude subscription or an Anthropic Console account. The desktop app requires a paid plan (Claude Code docs).
Read the full section
- Tool: Claude Code, according to the article URL. The model and version used are not stated in the reviewed excerpt.
- What Claude Code does: It is Anthropic's agentic tool. It reads and edits files, runs commands, connects to external data through MCP, and can run scheduled recurring tasks. It is available in the terminal, IDEs, a desktop app and the web (Claude Code docs).
- Access: Most surfaces need a Claude subscription or an Anthropic Console account. The desktop app requires a paid plan (Claude Code docs).
- Underlying approach: Earlier WSJ advice relied on chat tools plus exported history from IMDb or JustWatch (To Vima/WSJ).
Technical analysis for researchers and developers
The column's architecture is not documented. A coding agent can turn those into a reusable script or stored data, so the user doesn't have to repeat a chat prompt each time. The relevant documented building blocks are: Whether the column used any of these is unknown.
Read the full section
The column's architecture is not documented. Its stated criteria (ratings, episode count, excluded subjects) are structured filters. A coding agent can turn those into a reusable script or stored data, so the user doesn't have to repeat a chat prompt each time. The relevant documented building blocks are:
- persistent instructions in CLAUDE.md
- MCP connectors to outside data
- scheduled tasks that could refresh a watchlist (Claude Code docs)
Whether the column used any of these is unknown. Nothing about evaluation, data sources or code is published, so the setup can't be reproduced.
Claims and evidence
- Writer's experience: The writer reports that the tool reliably surfaces suitable series. This is first-person reporting with no measurement behind it (WSJ).
- Adoption figures: 41% use AI to find shows, and 57% of users rate it above search engines (RBR/Hub).
- Demand for episode count: 30% of viewers call episode or season count "very important" before starting a series, 40% among 16–34-year-olds (RBR/Hub).
Read the full section
- Writer's experience: The writer reports that the tool reliably surfaces suitable series. This is first-person reporting with no measurement behind it (WSJ).
- Adoption figures: 41% use AI to find shows, and 57% of users rate it above search engines (RBR/Hub). This is survey self-report, not a test of recommendation quality.
- Demand for episode count: 30% of viewers call episode or season count "very important" before starting a series, 40% among 16–34-year-olds (RBR/Hub). That supports the column's filter choice.
- Corroboration: No independent coverage of this specific column or setup was found.
Context and prior work
- Mood- and constraint-based prompting for media picks has been WSJ advice since 2024 (To Vima/WSJ).
- Platforms have responded with built-in options: Netflix's OpenAI-backed search lets users ask for things like "scary, but not too scary" (eWeek).
- The new element is the coding agent.
Read the full section
- Mood- and constraint-based prompting for media picks has been WSJ advice since 2024 (To Vima/WSJ).
- Platforms have responded with built-in options: Netflix's OpenAI-backed search lets users ask for things like "scary, but not too scary" (eWeek).
- The new element is the coding agent. It turns a one-off prompt into personal software, a shift Scientific American likened to what GarageBand did for music and Excel did for spreadsheets (Scientific American).
Limitations, safety and contested findings
- The full column was not available, so its methods, data sources and failure cases can't be assessed.
- Hub's data is mixed.
- Survey preference for AI over search is not evidence that the recommendations are accurate.
Read the full section
- The full column was not available, so its methods, data sources and failure cases can't be assessed.
- Hub's data is mixed. Personal recommendations and trailers still beat streaming algorithms for discovery, and 47% of viewers struggle to tell services' content apart (RBR/Hub).
- Survey preference for AI over search is not evidence that the recommendations are accurate.
- Scientific American's Claude Code piece is mostly the author's own experience. It also cites Anthropic's own claims about how much of Claude Code's code the tool writes, which are vendor statements (Scientific American).
Business and practitioner implications
- Streamers: Discovery is moving to general-purpose AI assistants outside their apps.
- Metadata providers: Episode counts, content warnings and ratings become key inputs for these filters.
- Developers: Small personal agents that keep stored preferences are a growing use case for consumer tools (Claude Code docs).
Read the full section
- Streamers: Discovery is moving to general-purpose AI assistants outside their apps. With 54% adoption among 16–34-year-olds, platforms risk losing control of the recommendation layer (RBR/Hub).
- Metadata providers: Episode counts, content warnings and ratings become key inputs for these filters. Clean, machine-readable data helps.
- Developers: Small personal agents that keep stored preferences are a growing use case for consumer tools (Claude Code docs).
The source trail.
Sources (7)
How AI Helps Me Find the Right TV Shows at the Right Times
Feed excerpt only.
wsj.com