Check Claude Brand Mention Context Before You Cite

By William Zhu · Cofounder, InfiniSynapse · Published: 2026-09-08 · Last updated: 2026-09-14 · Last verified: 2026-09-14 · Reviewed by the InfiniSynapse Data Team.

Reviewed by: InfiniSynapse Data Team · method review 2026-09-14. Anthropic intro 200, robots ClaudeBot Allow, aivisibility unauthorized. No Claude chat.

Trust / COI: About · Corrections. SEO Health is a commercial product from this team. Anthropic pages are citations, not a partnership.

TL;DR

Anthropic’s intro page answered 200. The aivisibility command still printed unauthorized. That pair is the Claude brand mention desk, not a chat.

Direct answer: Treat a Claude brand mention as one new thread × one recorded project or tool context × mention and citation coded apart. On 2026-09-14, https://docs.anthropic.com/en/docs/intro-to-claude answered HTTP 200 (396,958 bytes; title “Intro to Claude - Claude Platform Docs”). Privacy Policy answered 200. robots.txt allows ClaudeBot. seo-health aivisibility on a sibling printed 分析失败：unauthorized and exited 1. No Claude conversation was opened.

A Claude brand mention rate built from seeded diagnostic prompts that already name the brand is contamination. Start clean.

What you will learn

• Which Anthropic URLs we fetched
• Why ClaudeBot Allow is fetch policy
• Why unauthorized is not “Claude never mentioned us”
• Which context fields a later thread needs
• Where Gemini and alias ledgers live

What isolation records

Key definition: A Claude brand mention is an approved alias in Claude’s prose, coded after the prior-turn and tool context are written down. A citation is a source Claude itself exposed. A docs 200 is neither.

People open Claude brand mention after a teammate pastes a Project that already contains the brand handbook. That Project is a seed. Isolate it or discard the row.

Intro to Claude is the product intro we fetched. It is not a mention of InfiniSynapse. Anthropic Trust Center and Anthropic privacy policy are trust and privacy documents. They are not citations of our pages.

SEO Health remains the deterministic evidence foundation for robots and check reads. After a token exists, an InfiniSynapse Server API long-task workspace may consume SSE /api/ai/events for aivisibility; that job still does not open Claude.

Prompt injection stays a sibling risk

OWASP prompt injection is why we record whether tools or retrieved files were attached. A Claude brand mention that ignores attachments cannot explain a sudden alias hit.

UK NCSC secure AI development and CISA AI resources are public-sector frames. They do not code our spans.

Tools are context

Anthropic tool-use documentation describes tools. If a tool returns our homepage, that is retrieval context, not an unprompted mention. Write the tool name. Claude brand mention rows that hide tools are contaminated.

Public Claude-adjacent objects

User-agent InfiniSynapse-SH1-desk/2026-09-14.

Figure: Docs and robots succeed. The signed-in CLI does not. None of the bars is a Claude thread.

Anthropic transparency hub is a transparency index. NIST AI resource center is a resource hub. Neither is a mention log.

Robots is not a mention

ClaudeBot Allow means this host permits that user-agent to fetch. A Claude brand mention that treats Allow as “Claude already talks about us” is mixing crawlers with chats. The crawler job lives on the access siblings, not here.

How to reproduce the 200s and the Allow

curl -sS -A 'InfiniSynapse-SH1-desk/2026-09-14' -o /tmp/claude-intro.html \
  -w '%{http_code} %{size_download}\n' \
  https://docs.anthropic.com/en/docs/intro-to-claude

curl -sS -A 'InfiniSynapse-SH1-desk/2026-09-14' \
  https://infinisynapse.com/robots.txt | grep -n ClaudeBot

Expect a 200 and an Allow group unless those files change. Claude brand mention reproduction stops there until a reviewer stores a new thread offline.

Package infinitegrowth@0.1.1; -V printed 0.1.0. Do not spend a credit to “prove” a mention. Unauthorized is the honest line.

What a later Claude row still needs

New conversation. Project / tool inventory. Displayed model label or “not displayed.” Answer text. Exposed sources. Alias decision. Second reviewer on homonyms. Claude brand mention fails when a seeded Project is treated as organic discovery.

Mention versus citation

Prose alias = mention. Exposed source to an owned URL = citation. Keep both bits. A docs card can appear while the prose never names the brand.

Sample-size limit

Four public objects cannot establish how often Claude names the company, and they do not prove a handbook file caused a hit. As of 2026 this page is not a customer Claude export.

Failure modes that invent a thread

Intro 200 as a mention. The title is “Intro to Claude.”

Privacy 200 as a citation. It is Anthropic’s policy.

ClaudeBot Allow as a chat. Fetch policy.

Unauthorized as invisibility. Token missing.

Seeded Project as discovery. Context was planted.

Synthetic diagnostic rows. Older drafts invented them. They are gone.

A Claude brand mention memo that skips the robots versus CLI contrast is already late.

How a reviewer should read the four objects

Start with the intro 200. The title is “Intro to Claude.” That page explains Claude. It does not mention InfiniSynapse. File it as product documentation. A Claude brand mention ticket that pastes the intro URL as “Claude already knows us” is citing the wrong object.

Read Privacy Policy second. It is Anthropic’s policy. It is not a citation of our pages. Trust Center and transparency hub are the same class: vendor documents.

Read robots third. ClaudeBot Allow is fetch policy. The access siblings own crawler proof. This page only needs the Allow group so nobody confuses it with a chat. If ClaudeBot is later disallowed, write the new rule and the date. Do not convert that change into a mention rate.

Read unauthorized fourth. The sibling aivisibility run fetched a page and failed task create. That is a login gate. It is not Claude remaining silent. Do not spend a credit to “confirm” a mention unless the ticket is an InfiniSynapse task, and even then write that the task is not a Claude UI.

When a later thread exists, start a new conversation. Inventory Projects, tools, and uploaded files. If a tool returns our homepage, that is retrieval context. Code it as context, not as organic discovery. Claude brand mention rows that hide tools cannot be defended.

Save the visible answer before following a source. Count mention and citation independently. Homonyms go to a second reviewer. Seeded diagnostic prompts that already name the brand belong in a separate file labeled seeded.

If you only have the four objects, you have the desk. That is enough to reject a Claude-rate slide. It is not enough to claim Claude names the brand. Write both. A later thread log is an addendum. Do not put invented diagnostic rows back into the main table.

Two people curling the intro URL should see the same title class unless Anthropic moved docs. Two people opening Claude will not see identical prose. That difference is why the public objects are the desk.

Do not paste a Claude transcript into a public ticket. Redact. Store a hash. Claude brand mention that publishes raw threads to look empirical is leaking context.

What stays on disk after this desk

Keep the intro HTML, the privacy HTML, a robots excerpt that shows the ClaudeBot group, and the unauthorized stdout. That is the 2026-09-14 Claude brand mention packet. Threads, if they arrive later, go in threads/YYYY-MM-DD/ with context inventories.

Do not check in seeded diagnostics as if they were organic. Label seeded files seeded-. Claude brand mention that mixes seeded and clean rows in one average cannot explain a spike.

If robots later disallows ClaudeBot, store the new file beside the old Allow. That is an access change. It is not a mention change. Send that diff to the crawler siblings.

aivisibility remains optional and signed-in. If a later desk authorizes it, store taskId and usage and still write “not a Claude UI.” Do not rename that JSON to claude-mentions.json.

Redact transcripts. Store hashes of allowed excerpts. Public tickets get the four-object table, not the chat. Claude brand mention that pastes raw threads into Slack to look busy is a privacy incident, not empiricism.

The decision this page supports is isolation. It does not decide whether to buy a Claude tracker. When someone asks for a mention rate, point at the empty threads/ folder. Empty is allowed. A fake rate is not. That stop is the product of Claude brand mention here.

Package pin and binary version sit in versions.txt next to the unauthorized line so the two version strings stay attached to the gate.

Isolation checklist before anyone opens Claude

Write project names, custom instructions, and attached files on paper first. If the list is not empty, the first thread is seeded. Label it seeded or start a second, empty account. Claude brand mention that skips the paper list will argue later about why the brand appeared in turn one.

Tools that fetch the public web can retrieve our homepage even when the prompt never names us. That is context, not discovery. Code the tool. If you cannot name the tool, you cannot publish the row.

A second reviewer should read only the span, not the company handbook. If they cannot tell which entity was meant, mention stays unresolved. Unresolved is a legal state for Claude brand mention. Forcing yes to close a ticket is not.

When security asks whether we “tested Claude,” say: we fetched Anthropic docs and we read our robots Allow. We did not open a consumer thread. Stronger claims need the threads/ folder. Keep that sentence taped to the packet so Claude brand mention is not cited as a chat study.

Nearby guides

The Perplexity hub owns the Help 403. This page owns Claude-adjacent public objects.

The AI visibility hub remains the category page. Claude brand mention stays on isolation.

Frequently Asked Questions

Did we open a Claude conversation?

No. A Claude brand mention still has no answer text to code.

Does ClaudeBot Allow mean Claude mentioned us?

No. Allow is fetch policy. Mention is prose.

Is unauthorized a Claude finding?

No. The command never created a task. Login failed.

What counts as a Claude citation?

Only a source Claude itself exposed. A URL found later in Search is not that evidence.

May we publish a mention rate from this chart?

No. The chart is public objects versus exit 0. It cannot establish a rate.

Conclusion

Use Claude brand mention when the job is to refuse a chat score until context is isolated and a thread can be stored. This 2026-09-14 pass recorded two Anthropic 200s, a ClaudeBot Allow, and an unauthorized CLI. Stop when those four objects remain visible. A bounded public log beats an unsupported Claude percentage.

ClaudeBot Allow in robots is fetch policy; it is not a Claude brand mention. Until a clean thread exists, write unknown. If a later thread is stored, attach the paper isolation list to the first line of the file. If the list is missing, the thread is seeded even when the prose looks organic. Seeded rows can exist. They cannot sit in the same average as clean rows. Keep two folders. Keep two rates, or keep none. That split is the last rule this page can enforce without opening Claude itself. Reprint the unauthorized line whenever someone asks whether we “already ran visibility.” We ran a gate. We did not run a chat. Keep both sentences in the subject line of the next Claude brand mention ticket so the gate is not retold as a silent Claude. If the ticket still says “Claude said nothing,” it is the wrong ticket and should be rewritten before anyone opens a thread.