Final maintenance release in progress

AI-assisted review for
suspicious Gmail messages

Spam Likely provides a second opinion using an AI provider you choose. It can miss threats or flag legitimate mail and should never be your only security control.

BYOK Bring Your
Own API Keys
Manual You choose what to review
Direct Requests go to your provider
Final No ongoing maintenance planned
Features

A second opinion, with clear limits

AI can highlight suspicious wording, links, and social-engineering patterns. Its output can be incomplete or wrong, so verify important messages through trusted channels.

On-Demand Email Scanning

Click “Analyze” on a message you want reviewed. The extension sends selected message data to the AI provider you configured.

  • Manual review only in the final release
  • Explanations of possible warning signs
  • No automatic verdicts for your inbox
  • Not a guarantee of safety

Provider Choice

Configure one supported AI provider with your own API key. Provider privacy, retention, billing, and acceptable-use terms apply.

  • OpenAI, Anthropic, Gemini, or Mistral
  • One maintained model per provider
  • Keys remain in extension-local storage
  • No hosted Spam Likely analysis service

Bring Your Own Keys

Requests go directly from the extension to the provider you select rather than through a Spam Likely analysis server.

  • Your API keys, your control
  • No Spam Likely analysis server
  • Direct provider communication
  • Provider policies still apply

Gmail Integration

A small “Analyze” control appears in Gmail. The final release removes inbox monitoring and automatic scanning.

  • Native "Analyze" buttons
  • Visible success and failure states
  • No background inbox flagging
  • Non-intrusive design
How It Works

How manual review works

Review only the message you select, then treat the result as one input to your own judgment.

1

Manual Scanning

Open the relevant message and click “Analyze.” The final release selects the currently expanded visible message.

  • Click "Analyze Email" button
  • Sender, subject, current message text, and links are sent
  • Attachments and quoted thread history are excluded
  • The current-message text is length-limited
2

Review the second opinion

A low-risk response means only that the model found no suspicious indicators. Verify requests involving money, credentials, or sensitive information independently.

  • Never rely on the result as a guarantee
  • Use Gmail and organizational security controls
  • Confirm unusual requests through another channel
  • Report phishing through Gmail when appropriate
Architecture

Your Data, Your Choice

The final release is manual and BYOK. It does not require mailbox credentials or route analysis through a Spam Likely server.

Feature Spam Likely Browser Extension Hosted Services
Email Password Storage
Never stored
Varies by service
Email Selection
You choose what to analyze
Varies by configuration
Data Processing
Direct to AI provider you choose
Depends on service architecture
AI Provider Choice
Choose one supported provider
Provider selected by the service
Security Risk
Limited local metadata only
An additional service processes data
User Control
Manual review only
Varies by service
Project status

Final maintenance release

Spam Likely is being retired. Existing users can receive one final cleanup and manual-review update before the Store listing is removed.

Learn the warning signs

The project’s phishing-awareness articles remain available as educational material.