Telegram Keyword Filters Match Too Much? Use Whole-Word Rules

If a Telegram keyword rule catches longer words you meant to leave alone, define the word boundary before adding more exclusions. In RedFox Auto Forward Telegram, use a Regex whitelist to admit a whole word, or a Regex blacklist to reject it. For a simple lowercase English example, \bsale\b targets sale without also targeting wholesale. Test the rule on one controlled task with both wanted and unwanted examples before changing a live feed.

This guide is for an existing Telegram-to-Telegram route that works, but whose text rules select too much or too little. The goal is a precise eligibility rule—not text deletion, historical recovery or a new forwarding setup.

Product identity: RedFox Auto Forward Telegram uses @AutoForwardNew_Bot. The previous bot has moved to this current handle. Auto-Bot.io is an independent editorial publisher, not the developer or operator of the product.

The featured artwork is a conceptual illustration of whole-word matching, not a Telegram or Auto Forward screenshot. The examples below are a proposed test set based on the documented controls; they are not results from a live forwarding test.

Decide whether you want to select a post or change its text

Start with a plain-English policy. “Forward a post when it contains the word sale” and “remove the word sale from a post” are different jobs. The official Whitelist guide documents content-based admission with Basic and Regex rules. The Blacklist guide documents blocking messages that match a rule.

• Whitelist: the message must satisfy the admission rule. For this guide’s example, admit a post containing the whole word sale.
• Blacklist: a matching message is blocked. Use the same word rule when your policy is to exclude posts containing sale.
• Content modification: if the post should arrive with different wording, investigate Replace or Cleaner instead. A blacklist is not an instruction to erase one word and forward the rest.

Write down which of those outcomes you need. A correct pattern in the wrong rule type can produce the opposite feed. For a post that arrives with missing photo content or captions, use the separate photo filters versus Cleaner diagnostic rather than treating every missing part as a word-match failure.

Use a whole-word rule for a simple English keyword

The official Blacklist FAQ explicitly recommends \bword\b for a full-word match and says to enable Regex Mode for regex patterns. Applying that documented form to the word sale gives:

\bsale\b

Enter that expression in a Regex rule. Do not paste it into Basic mode and assume the backslashes will be interpreted as regex. This example is a pattern to validate in your task, not a promise about every language, character set or account configuration.

For the lowercase ASCII examples here, the boundary separates sale from neighboring word characters. That lets sale today differ from wholesale today. It also means punctuation can be significant: a word boundary does not mean “spaces only.” If a hashtag or hyphen should change eligibility, include that exact spelling in the test set.

Do not diagnose the product’s Basic matching behavior from a rule name alone. Record the exact existing pattern, mode and a source message that produced the surprising result. The useful correction is the smallest rule that matches your intended policy, not an assumption that every Basic rule behaves the same way.

Build a test set that can disprove your rule

Use a source and a private destination you control, containing only harmless content you may redistribute. Preserve the current settings first. If you need a separate test route, follow the existing setup guide; do not rebuild a working production route just to inspect one pattern.

The following are expected pattern checks for \bsale\b, not observed delivery results. “Match” means the expression finds the intended word. A whitelist uses that match to admit; a blacklist uses it to block. Other active task rules can still affect final delivery.

Now add policy probes: SALE today, #sale, pre-sale and one real formatted source post. Decide beforehand which should pass your editorial policy, then record what the task actually does. The lowercase expression does not establish case-insensitive matching; neither the docs’ whole-word tip nor an external regex tester establishes the service’s complete engine settings.

If #sale matches but your policy excludes hashtags, the expression may be working exactly as a boundary expression should. Your policy needs a more specific pattern. Do not add an arbitrary blacklist entry to hide a mismatch you have not explained.

Apply the rule to one task—not every route

Both official rule guides distinguish task-scoped activation from Apply For All Tasks. Use that distinction to keep a local investigation from changing unrelated feeds.

1. Record the baseline. Save the task label, rule label, Basic or Regex mode, exact pattern, Post or Reply scope, and active whitelist/blacklist associations. Check which other tasks use each existing rule, and keep a private copy of the previous values and associations for rollback.
2. Create a separate, clearly named test rule. Open Whitelist or Blacklist from the dashboard, choose Add New, select Regex, and enter \bsale\b. Use a new label such as sale_whole_word_test and associate this new rule only with the controlled task. The guides list Edit separately from Activate/Deactivate For Task: task-scoped activation does not establish that editing a reusable rule is isolated. Edit an existing rule only after verifying that no other task uses it; otherwise leave it unchanged and use the new test rule.
3. Choose the intended scope. For this first test, use ordinary new posts and the Post scope. The guides list Post and Reply separately; a successful post test does not validate reply behavior.
4. Associate only the controlled task. Use Apply For Task or the rule’s Activate/Deactivate For Task control as appropriate. Do not choose Apply For All Tasks to solve a one-route problem.
5. Send the fresh specimens. Run the expected matches, near misses and policy probes. Compare the source text with the exact private destination, noting whether the whole post arrived and any visible error.
6. Keep or restore one change. If the result does not explain the symptom, restore only the controlled task’s previous association and remove its test-rule association before testing a different hypothesis. Leave every other task and shared rule unchanged. Do not stack unexplained edits.

Keep unrelated transformations out of the first controlled comparison where possible. That is a diagnostic simplification, not a claim about Auto Forward Telegram’s internal processing order or about which rule wins when conditions conflict.

Check entitlement before configuring the test. The current product site, checked October 16, 2026, separates basic Free routing from paid advanced controls and lists allow/block lists under Diamond. Eligible new users may have Trial Platinum. Confirm the active plan and available controls in your account; basic Free routing is not a promise that this Regex workflow is included.

If the visible word looks right, inspect the actual source text

A visually similar word is not necessarily the same sequence of characters. Copy the real source text rather than retyping a cleaner version from memory. Compare punctuation, capitalization, line breaks, underscores and text formatting before broadening the expression.

RedFox’s special-formatting guide specifically addresses Replace, Whitelist and Blacklist mismatches and describes inspecting raw content or HTML through FindMyIDs Bot. Use only a harmless owned specimen with any helper bot; do not forward confidential customer messages or account material merely to inspect formatting.

That guide is a reason to investigate the actual representation, not proof that bold text always fails. Likewise, a keyword printed inside an image is not automatically selectable message text. Keep image recognition, caption matching and plain text matching as separate questions.

For multilingual keywords or punctuation-heavy identifiers, do not treat this English \b example as a universal recipe. Write down the exact characters that should count as a boundary and test representative real samples. If the observed behavior remains unexplained, keep the production rule unchanged and ask product support about that exact pattern and sample.

Use the result to narrow the rule—not to replay history

For each sample, retain a short private record: source wording, posting time, task label, rule type, exact pattern, intended decision, observed whole-post delivery and any error. Separate pattern match from final delivery. A neutral post disappearing under a whitelist can be expected; a known matching post failing to arrive may involve another active rule or a route error.

If a controlled test explains the problem, preserve every other task’s rule associations and change only the intended live task’s association to a newly created rule exclusive to that task. Edit an existing rule only after verifying that no other task uses it; do not repurpose a shared rule. Check a fresh wanted post and a fresh near miss after that change. Keep the previous pattern and task association available for rollback. Do not assume a new filter change recovers previously skipped messages, and do not clone source history to prove a new-message rule.

If support is needed, provide the exact pattern and mode, expected versus observed behavior, task scope and a redacted harmless specimen. Do not include login codes, passwords or private channel content. A screenshot of a rule label without its pattern and scope is not enough to reproduce the decision.

For a different selection problem—combining a sender condition with a keyword or filtering replies by context—read the advanced filtering guide. Whole-word precision answers a narrower question: what counts as the keyword inside this message?

Inspect one keyword rule in Auto Forward Telegram

Open the RedFox Web App, select a controlled task and compare its current rule with the policy you wrote down. Start with one keyword, one whole-word expression and both positive and negative samples. Expand only after the actual result makes sense.

You can also manage the product through its official iOS app or Android app. Web/App is the recommended path for fuller management than the limited Telegram-bot entrypoint; active tasks run in RedFox’s cloud without keeping the management interface open. See Auto Forward Telegram for the supported routing overview.

FAQ

Should I put the whole-word expression in Whitelist or Blacklist?

Use a Regex whitelist when a matching post should be eligible to forward. Use a Regex blacklist when it should be rejected. Neither choice means “remove only the matched word.” Other active task rules still matter.

Can I enter backslash-b in Basic mode?

The official FAQ says to enable Regex Mode for these patterns. Choose Regex for \bsale\b; do not assume Basic interprets it as an expression.

Does this also match uppercase or every language?

Do not assume that. This guide uses a lowercase English example. Test uppercase, real punctuation and the exact language you publish in before using the rule on a live destination.

What if a matching new post still does not arrive?

Recheck the exact task association, Post/Reply scope, other active rules and any route error. Preserve the specimen and observed result rather than disabling every filter or repeatedly sending to a real audience.