Dotdash Meredith / People Inc Agentic Access

generated: '2026-08-12'
method: probed
source: live HTTP probes of people.inc and the brand estate, 2026-08-12
description: |
  People Inc has no authenticated developer API, so its "agentic access"
  contract is not an x-agentic-access block over OpenAPI operations — it is a
  DENY-AND-PRICE policy enforced at the CDN edge over the whole content
  estate. This artifact records that contract as it was actually observed on
  the wire, not as it is described in marketing.

  The headline finding: every People Inc host answers HTTP 402 Payment
  Required — with a licensing contact in the body — to a request that
  identifies itself as ClaudeBot. This is a machine-readable commercial
  signal, and it is the single clearest statement of the company's agent
  posture anywhere on its public surface.

posture:
  stance: deny-then-license
  enforcement: edge
  edge_vendor: Cloudflare
  robots_policy: comprehensive-ai-denylist
  paid_crawl_signal: true
  licensing_contact: contentlicensing@people.inc
  summary: |
    Editorial content is closed to AI crawlers by robots.txt AND by an edge
    rule that returns 402. Discovery metadata (robots.txt, sitemaps,
    security.txt) stays open to everyone. Sponsored/branded content is
    explicitly carved OUT of the AI denylist and left crawlable.

observed_responses:

  - probe: HTTP GET with User-Agent "ClaudeBot/1.0 (+https://www.anthropic.com/claude-bot)"
    hosts:
      - url: https://www.people.inc/news-awards
        status: 402
        content_type: text/plain
        bytes: 109
      - url: https://people.com/
        status: 402
        content_type: text/plain
        bytes: 109
      - url: https://www.allrecipes.com/
        status: 402
        content_type: text/plain
        bytes: 109
      - url: https://www.investopedia.com/
        status: 402
        content_type: text/plain
        bytes: 109
    body: |
      Payment Required - If you wish to license content from People Inc, please contact contentlicensing@people.inc
    response_headers_of_note:
      server: cloudflare
      cache-control: private, max-age=0, no-store, no-cache, must-revalidate
      strict-transport-security: max-age=15552000
    reading: |
      402 is the correct and honest status for this: not "you are forbidden"
      but "this costs money, here is who to ask". It is a priced boundary, and
      an agent can act on it. Very few publishers in the catalog return 402 at
      all; most return 403 or a soft-200 paywall shell that an agent cannot
      distinguish from content.

  - probe: HTTP GET with User-Agent "GPTBot/1.2"
    hosts:
      - url: https://www.people.inc/news-awards
        status: 403
        content_type: text/html
        bytes: 680204
    reading: |
      DIFFERENT AGENTS GET DIFFERENT ANSWERS. GPTBot is refused with the
      generic bot-management interstitial (403 + ~680KB of HTML), not the
      402 licensing offer. Only the Anthropic crawler received the priced
      response in this probe. Whether that reflects a per-vendor rule or a
      different rule ordering is not determinable from outside; what is
      recorded here is the observed divergence.

  - probe: HTTP GET with a desktop Chrome user-agent over curl
    hosts:
      - url: https://www.people.inc/news-awards
        status: 403
        content_type: text/html
        bytes: 680482
      - url: https://people.com/
        status: 403
        content_type: text/html
      - url: https://people.com/feed
        status: 403
        content_type: text/html
    reading: |
      A browser user-agent is NOT sufficient. The edge fingerprints the client
      beyond the UA string, so every non-browser client is refused on HTML
      paths regardless of what it claims to be. This is why the RSS feed
      endpoints recorded in apis.yml cannot be verified from a script.

  - probe: allowlisted machine paths, any user-agent including ClaudeBot
    hosts:
      - url: https://people.com/robots.txt
        status: 200
        content_type: text/plain
      - url: https://people.com/sitemap.xml
        status: 200
        content_type: text/xml
      - url: https://people.com/google-news-sitemap.xml
        status: 200
        content_type: text/xml
      - url: https://people.com/.well-known/security.txt
        status: 200
        content_type: text/plain
    reading: |
      The discovery layer is deliberately left open to everyone, including the
      crawlers the same edge charges for content. People Inc is closing the
      corpus, not the map.

robots_policy:
  source: https://people.com/robots.txt
  status: 200
  fetched: '2026-08-12'
  user_agent_directives: 78
  license_notice_in_comments: true
  license_notice: |
    People Inc. content is made available for your non-commercial use subject
    to Terms of Use at https://www.people.inc/brands-termsofservice. Use of
    any crawler to data mine or scrape for any purpose other than directing
    traffic or serving authorized advertisements is prohibited without prior
    written permission. Prohibited uses expressly include: (1) text and data
    mining under Art. 4 of the EU Directive on Copyright in the Digital Single
    Market; (2) development or operation of any AI, ML, or LLM technology,
    including training, fine-tuning, or retrieval-augmented generation; and
    (3) creating data sets containing People Inc. content or sharing it.
    Contact contentlicensing@people.inc.
  legal_hooks:
    - EU DSM Directive Art. 4 TDM reservation (machine-readable opt-out)
    - explicit RAG prohibition
    - explicit dataset-creation prohibition
  fully_denied_examples:
    - Google-Extended
    - anthropic-ai
    - Claude-SearchBot
    - ClaudeBot
    - Claude-User
    - Claude-Web
    - CCBot
    - cohere-ai
    - cohere-training-data-crawler
    - Meta-ExternalAgent
    - Meta-ExternalFetcher
    - meta-webindexer
    - PerplexityBot
    - Perplexity-User
    - Bytespider
    - Applebot-Extended
    - Amazonbot
    - AmazonBuyForMe
    - Amzn-SearchBot
    - MistralAI-Index
    - MistralAI-user
    - DuckAssistBot
    - Google-CloudVertexBot
    - FirecrawlAgent
    - ImagesiftBot
    - Kagibot
    - YouBot
    - iAskBot
    - Quora-Bot
    - Webzio-Extended
  partial_allow:
    - user_agents: [GPTBot, OAI-SearchBot, ChatGPT-User]
      rule: 'Disallow: /thmb/'
      reading: |
        The OpenAI crawlers are the ONE family not blanket-denied in
        robots.txt — only the /thmb/ image-thumbnail path is disallowed.
        People Inc has a commercial relationship with OpenAI; this asymmetry
        is the robots.txt expression of it.
    - user_agents: [Pinterest, Pinterestbot]
      rule: 'Disallow: (empty value — full allow)'
    - user_agents: [AmazonAdBot]
      rule: 'Allow: /'
  sponsored_content_carve_out:
    rule: |
      Under the block that denies Google-Extended, anthropic-ai, CCBot,
      Claude-SearchBot, cohere-ai, Meta-ExternalAgent, meta-webindexer and
      PerplexityBot with "Disallow: /", two Allow lines reopen specific paths:
        Allow: */presented/
        Allow: */integrated/
    reading: |
      SPONSORED AND BRANDED CONTENT IS EXEMPTED FROM THE AI BLOCK. The paths
      People Inc keeps open to the AI crawlers it otherwise denies are the
      paths that carry advertiser-paid content. Editorial is withheld;
      advertising is offered. This is the clearest instance of agent-facing
      commercial asymmetry in the media segment of the catalog and it is
      stated in the provider's own published file, not inferred.

corporate_host_policy:
  source: https://www.people.inc/robots.txt
  status: 200
  fetched: '2026-08-12'
  note: |
    The corporate host runs a SEPARATE, differently-authored robots.txt with
    section-header comments ("AI TRAINING & CONTENT SCRAPING BOTS - BLOCKED").
    It includes two wildcard user-agent patterns — "User-agent: *Claude*" and
    "User-agent: *AI*" — which are NOT valid Robots Exclusion Protocol; RFC
    9309 user-agent matching is a prefix match on a token, with no glob
    support. Those two groups match nothing and enforce nothing. The 402 edge
    rule is what actually holds this host closed.
  invalid_directives:
    - directive: 'User-agent: *Claude*'
      reason: RFC 9309 does not support glob patterns in user-agent tokens.
    - directive: 'User-agent: *AI*'
      reason: RFC 9309 does not support glob patterns in user-agent tokens.

consent_signals:
  aipref: false
  content_signals: false
  tdm_reservation_protocol: false
  web_bot_auth: false
  http_message_signatures: false
  robots_txt_tdm_reservation: true
  note: |
    The TDM opt-out is asserted in robots.txt PROSE (a comment block naming
    EU DSM Art. 4), not in any of the machine-parseable consent standards.
    No /.well-known/ consent document, no Content-Signal header, no AIPREF
    vocabulary. An agent must read English to learn the policy, or read a 402.

agent_readiness_reading:
  callable_surface: false
  priced_refusal: true
  discovery_open: true
  summary: |
    There is nothing here for an agent to CALL. What there is, is unusually
    well-formed refusal: a documented denylist, a priced 402 with a named
    counterparty, and an open discovery layer. For a publisher that is a
    coherent posture, and it is more legible to an agent than most publishers
    who simply return 403 or a soft-200 paywall shell.

x-evidence:
  fetched: '2026-08-12'
  method: curl with varied User-Agent strings; no credentials used
  user_agents_tested:
    - ClaudeBot/1.0 (+https://www.anthropic.com/claude-bot)
    - GPTBot/1.2
    - Mozilla/5.0 ... Chrome/124.0 Safari/537.36
    - curl/8.7.1
    - Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)