The SFR antispam filter operates on multiple layers: a protocol analysis (SPF, DKIM, DMARC), content scoring, and recently, a security engine based on artificial intelligence capable of automatically quarantining an account. Understanding these mechanisms allows you to adjust your sends to reach the inbox rather than the spam folder.
AI Quarantine at SFR: a filtering system that monitors account behavior
At SFR, filtering does not stop at the content of the email. An additional layer analyzes the behavior of the sending account itself. An updated SFR Mail guide in 2026 describes a security engine that triggers automatic quarantine when it detects suspicious activity.
An SFR account that suddenly jumps from a few emails per day to several dozen marketing messages will be temporarily restricted. Sending may be blocked or filtering tightened, even if the content of the emails is impeccable.
To avoid this scenario, it is essential to work on the reputation of the SFR account itself: regular and gradual usage, absence of logins from unusual locations, and slow volume increases. This approach complements the detailed recommendations on the SFR antispam filter by integrating the behavioral dimension that most guides overlook.
SPF, DKIM, and DMARC Authentication: the technical foundation for sending to SFR
Since the 2024-2026 period, French internet service providers (including SFR) have implemented authentication requirements comparable to those of Gmail. Three protocols are now systematically verified: SPF, DKIM, and DMARC.
SPF: declaring authorized servers
The SPF record, published in your domain’s DNS, lists the IP addresses authorized to send emails on your behalf. If your message is sent from an undeclared server, SFR may reject it as soon as the SMTP connection is made.
DKIM: cryptographically signing each message
DKIM adds a digital signature to the email header. The SFR server verifies this signature by comparing it to the public key published in your DNS. An email without a valid DKIM loses credibility with the filter immediately.
DMARC: aligning SPF and DKIM under an explicit policy
DMARC tells the receiving server what to do if SPF or DKIM fails (accept, quarantine, reject). It also sends reports that help detect impersonations of your domain. A properly configured DMARC policy significantly reduces rejections at SFR.
Email content and triggering signals for the SFR filter
With authentication set, the content of the message remains a determining scoring factor. The SFR filter analyzes several elements before classifying an email as spam.
- The subject line: excessive capitalization, repeated exclamation points, and aggressive commercial terms (“free,” “exclusive offer,” “urgent”) increase the spam score. A descriptive and sober subject line performs better.
- The image/text ratio: an email primarily composed of images with little text is a classic spam signal. Prioritize textual content with complementary images.
- Links in the body: an excessive number of links or non-transparent shortened URLs trigger an alert. Each link should point to a domain consistent with the sender.
- The unsubscribe link: its absence is now a deal-breaker. French ISPs require a one-click unsubscribe mechanism, aligned with standards imposed since 2024.
A often overlooked point: tracking pixels (invisible pixels embedded to measure openings). The CNIL regulates their use, and SFR filters may degrade the score of an email that contains several or whose pixel comes from a poorly reputed domain.
Sender reputation and complaint rate management
The reputation of the sending IP address and domain weighs heavily in the SFR filter’s decision. This reputation is built over time and degrades very quickly.
The complaint rate is the most destructive signal. Each time an SFR recipient clicks “Report as spam,” the ISP records this action. Beyond a certain threshold, subsequent emails are systematically redirected to the spam folder or even rejected at the SMTP level with a 554 error.
To maintain a low complaint rate, three levers work:
- Only send to recipients who have explicitly consented (confirmed opt-in). Purchased lists mechanically generate high complaints.
- Regularly clean the database: remove inactive addresses, permanent bounces, and spam traps that degrade the domain’s reputation.
- Maintain a sending frequency consistent with the recipients’ expectations. A sudden spike in volume is interpreted as suspicious behavior, both by SFR’s AI engine and by traditional filters.
SMTP port and server configuration: a detail that blocks sending
A common technical issue for SFR users concerns the SMTP sending port. Port 25, historically used for outgoing mail, is increasingly blocked by ISPs and hosting providers to limit outgoing spam.
SFR recommends using port 587 with STARTTLS encryption for authenticated sending. SFR’s SMTP settings (server smtp.sfr.fr) require authentication with a username and password. A send configured on port 25 without authentication will likely be rejected or marked as suspicious.
This constraint also applies to third-party emailing software: ensuring that the SMTP connection uses the correct port and encryption protocol avoids silent rejections that many mistakenly attribute to the content of the message.
The SFR AI filter adds a behavioral dimension that makes gradual scaling and the cleanliness of the recipient database more decisive than the email drafting itself. A properly authenticated send (SPF, DKIM, DMARC), addressed to active contacts, and sent from an account with a stable history, passes through these filters without difficulty.