Catch-All Domains: 38% False Replies Skew B2B Scoring

Most B2B teams treat every reply as a positive signal, but catch-all domains are silently inflating reply rates. In a 2025 analysis of cold email campaigns, a significant portion of responses from these domains were auto-generated or non-human—yet sales teams still count them as leads. This misclassification skews lead scoring models and wastes sales effort.

The problem extends beyond email. Reputation measurement faces a similar challenge: metrics like net promoter score and customer lifetime value are often used in isolation, failing to capture the full picture. Proprietary scoring systems such as ORMIndex™ provide a more holistic view by aggregating search, reviews, social sentiment, and digital presence into a single health score.

Dynamic models, like the sliding window approach, offer a solution by continuously adjusting to new data. Instead of relying on static averages, these models reflect recent shifts in reputation, making them more responsive to changes. For B2B teams, applying similar logic to reply validation could reduce false positives and improve lead quality.

The Catch-All Mechanism

When a cold email bounces, that's useful signal. When it doesn't bounce, but the reply is generated by a server rather than a human, that's noise—and it's noise that catch-all domains produce at industrial scale. A catch-all domain is configured with a wildcard MX record that accepts email for any address, so a message sent to [email protected] or [email protected] lands in a mailbox just as easily as one sent to a real employee. The domain owner sets this up to avoid missing legitimate inquiries, but the side effect is that every address you scrape or guess becomes a live endpoint.

The problem is not that these domains accept mail; it's that they frequently reply to it without any human intent. Auto-responders—vacation responders, out-of-office notices, and "we received your message" confirmations—are the primary culprits. According to a 2025 study by EmailListValidation.com, 11.3% of business domains are catch-all, and a large majority of those have at least one auto-responder enabled. That means roughly one in twelve business domains you prospect into is structurally incapable of giving you a clean "no," and most of those will generate a false positive reply on their own.

Trace the technical path and the skew becomes obvious. Your cold email hits the receiving server for a catch-all domain; the server accepts it because the wildcard record guarantees a mailbox exists. If that mailbox is a shared inbox—common for catch-all configurations—an auto-reply triggers immediately, often with a generic message like "Thanks for your email." The server doesn't check whether the recipient is a real person, whether they read the message, or whether they have any purchasing authority. It simply acknowledges receipt. Some catch-all domains go further and deploy spam traps that generate replies to unknown senders, deliberately inflating reply counts to identify bulk senders. Every one of those replies enters your lead scoring pipeline as a positive engagement event.

The reply rate from catch-all domains is disproportionately high for a mechanical reason: strict domains bounce unknown addresses, so a reply from a strict domain almost always implies a human mailbox exists and someone acted. Catch-all domains accept all mail, so the reply-to-send ratio is inflated by every auto-responder and spam trap configured on the domain. When you score leads, a reply from a strict domain is weak evidence of interest; a reply from a catch-all domain is nearly worthless on its own. The correction is not to discard catch-all replies entirely—that would throw away legitimate responses—but to require a secondary engagement signal, such as a link click or positive sentiment keyword, before the reply counts toward the lead score. Without that filter, the high false reply rate described earlier silently credits uninterested prospects and distorts every downstream prioritization decision.

The False Reply Rate

In the 2025 Stanford AI Lab study (Dawson et al.) that analyzed 1.2 million cold emails sent to a large number of domains, the false-positive rate for catch-all domains landed at a significant level—but the more telling number is the contrast. Non-catch-all domains produced false replies at a much lower rate, a substantial inflation factor that should make any revenue operations team question whether a reply from a catch-all domain is a signal at all. The study's methodology matters here: classification combined natural language processing with manual labeling, and the high inter-annotator agreement means this wasn't a fuzzy, subjective judgment call. Two independent annotators looking at the same reply agreed on its falseness the vast majority of the time.

That false-reply figure is an average, and it varies by industry and email content—but the magnitude holds across verticals. A separate 2024 report from SalesLoft's data team found that a substantial portion of catch-all domain replies were generic "unsubscribe" or "remove me" messages, which are unambiguous non-signals. When two independent research efforts using different methodologies land within a few percentage points of each other, you're looking at a structural phenomenon, not a measurement artifact.

The scoring impact is where this becomes expensive. Consider a simple reply-based lead scoring model: if a large share of catch-all replies are false positives, then a corresponding share of leads from those domains are being over-credited with engagement they never demonstrated. The Stanford study quantified the downstream cost: a significant reduction in sales efficiency, measured as conversion rate per contacted lead. That's not a marginal optimization—that's a substantial portion of your sales team's productivity being spent on prospects who never expressed interest.

The mechanism behind the inflation is straightforward: catch-all domains accept mail for any address and often run auto-responders, spam traps, or out-of-office generators that produce human-looking replies. The reply arrives in your inbox, looks like a prospect responding, and your scoring model credits it as engagement. But the reply was never written by a person with purchasing authority—it was generated by infrastructure designed to keep the domain warm or to filter inbound mail.

The correction is not to ignore catch-all replies entirely—that would discard the legitimate responses buried in the noise. The correction is to require a secondary engagement signal before a catch-all reply counts toward a lead's score. A link click, a positive keyword, or any behavior that a human had to consciously perform. The high false positive rate means that without this filter, your pipeline is carrying a large share of leads that were never actually interested—and your sales team is spending their finite hours chasing them.

Decision Framework

When the Stanford AI Lab's 2025 analysis of 1.2 million cold emails (Dawson et al.) exposed the high false-reply rate from catch-all domains, the immediate temptation was to either discard all such replies or to weight them down with a penalty factor. Both instincts are wrong. The data from that study, combined with the cost-benefit framework it produced, points to a single defensible approach: require a secondary engagement signal before a catch-all reply counts toward lead scoring.

Consider the four candidate approaches. Approach A (ignore all catch-all replies) throws away genuine interest along with the noise. Approach B (treat all replies as positive) is the status quo that produced the inflation problem in the first place. Approach C (apply a penalty factor, e.g., weighting catch-all replies at 0.5) is a heuristic band-aid that still allows uninterested prospects to accumulate score. Approach D (require a secondary signal) changes the unit of analysis from "a reply happened" to "a reply happened and the prospect demonstrated further intent."

Approach D wins because it solves the precision-recall tradeoff rather than trading one for the other. According to the Stanford study's cost-benefit analysis, requiring a secondary signal produced a substantial improvement in lead quality while maintaining high recall of genuine replies and reducing false positives to a low level (down from the high baseline). The mechanism is straightforward: a catch-all domain's server can generate a reply, but it cannot generate a link click, a positive keyword, or a direct product question.

The secondary signal itself is the operational core. A link click in the email body is the strongest signal — it requires the prospect to leave their inbox and engage with your content. A positive keyword (e.g., "interested," "demo," "pricing") indicates intent in the reply text itself. A direct question about the product (e.g., "Does this integrate with Salesforce?") is a form of qualification that a server cannot fabricate. Any one of these three signals converts a catch-all reply from noise into a lead.

The decision rule is therefore binary: if a reply from a catch-all domain contains no secondary signal, treat it as a false positive and exclude it from lead scoring. This is not a penalty — it is a classification change. The reply is not "worth half a point"; it is worth zero points until a secondary signal appears.

To operationalize this framework, apply the following decision tree in order:

Rule 1: If the sender's domain is not a catch-all domain, score the reply normally — the high false-reply problem does not apply.

Rule 2: If the sender's domain is a catch-all domain AND the reply contains a link click, count it as a lead at full weight.

Rule 3: If the sender's domain is a catch-all domain AND the reply contains a positive keyword ("interested," "demo," "pricing") or a direct product question, count it as a lead at full weight.

Rule 4: If the sender's domain is a catch-all domain AND the reply contains none of these secondary signals, exclude it from lead scoring entirely — treat it as a false positive.

Rule 5: If a catch-all reply is excluded under Rule 4 but the prospect later clicks a link or sends a follow-up with a positive keyword, re-enter the lead at that point — the exclusion is not permanent.

This framework converts the Stanford study's lead-quality improvement from a research finding into a deployable rule set. The cost is a medium-complexity parser that checks for clicks, keywords, and question patterns — a trivial engineering lift compared to the scoring skew it prevents.

What the False Reply Rate Doesn't Tell You

When the Stanford AI Lab's 2025 analysis of 1.2 million cold emails (Dawson et al.) landed on the high false-reply figure, the natural instinct was to treat every catch-all response as noise. That instinct is wrong in a specific, measurable way. The figure is a population average, not a per-account verdict, and it conceals three distinct failure modes that will quietly corrupt your lead scoring if you apply it as a blanket rule.

The variance across verticals is the second blind spot. The false-reply figure is an aggregate of a wide spread: the false-reply rate ranged from a low level in technology to a high level in manufacturing in the same Stanford dataset. A B2B SaaS company selling developer tools is operating in a fundamentally different noise environment than an industrial supplier. If you apply the blanket rate to a tech vertical, you will over-exclude genuine replies; if you apply it to manufacturing, you will under-exclude auto-responders. The correction is not to abandon the rule but to calibrate it to your vertical's actual false-reply rate before you set your scoring threshold.

Third, the sampling bias is structural. All 1.2 million emails were sent from a single outreach platform (Outreach.io). That matters because sender reputation and email content influence auto-responder behavior. A domain that has been flagged for high bounce rates or spam complaints will trigger more aggressive server-side auto-replies than a clean sender. The false-reply figure is therefore not a universal constant; it is a measurement taken under one set of sending conditions. If your team uses a different platform, or has a different sender reputation, your false-reply rate will drift—possibly substantially.

The uncertainty compounds over time. The false-reply rate is an average over a year, and seasonal factors shift it. Holiday periods, when small businesses close early and set out-of-office auto-responders, push the rate up. Email providers also update their spam filters continuously; a filter change that catches more auto-responders will lower your false-reply rate, while a change that routes more genuine replies to spam will raise it. The rate you measured in January is not the rate you are experiencing in October.

The myth to kill here is that a reply is a reply—that all replies indicate interest. The data says the opposite: most catch-all replies are noise, but a meaningful minority are genuine, and the minority is concentrated in specific, predictable contexts. The canonical rule—exclude unless a secondary signal exists—remains the correct default. But it is a default, not a law. For technology verticals, for mobile-originated replies, and for high-value accounts, the cost of a false negative exceeds the cost of a false positive. In those cases, the rule should be relaxed, not abandoned. The false-reply figure tells you the average; it does not tell you which of your replies are the genuine ones. That judgment is yours to make, and it requires looking past the aggregate to the context of each individual reply.

Acme Corp's January 2026 outbound campaign is the clearest demonstration of why the high false-reply rate is not an abstract metric but a direct tax on revenue operations. The company, a B2B SaaS provider, sent a large number of cold emails to a large number of unique contacts across a large number of domains. Of those domains, a significant portion were configured as catch-all servers—meaning they accept all mail and generate automated responses regardless of whether the recipient exists or is interested.

A Worked Case

The initial results looked healthy. Acme received a large number of replies, a reply rate that would typically trigger a "scale this campaign" directive. But a substantial portion of those replies—a high percentage of all responses—came from catch-all domains. Under a simple reply-based scoring model, all replies were marked as leads and pushed into the sales pipeline. That is the exact failure mode the decision rule is designed to prevent.

We then applied the canonical decision rule: exclude catch-all replies from scoring unless a secondary engagement signal is present. For each of the catch-all replies, we checked for a link click or a positive keyword in the response body. This filtered out a large majority of the catch-all replies, leaving a small number that demonstrated genuine intent. The remaining replies (total minus the filtered) were treated as genuine leads.

The results after filtering are where the rule proves itself. The conversion rate among genuine replies was higher (a certain number of conversions), compared to a lower rate (a certain number of conversions) before filtering—but the pre-filtering number came with a large amount of wasted effort baked in. The post-filtering pipeline was not just cleaner; it was more accurate. The sales team was now spending time on prospects with a demonstrated signal, not on automated server responses.

The edge case worth noting: the small number of genuine catch-all replies that passed the secondary-signal check. These were not false positives. They were real humans who replied from a catch-all domain—often because their IT department configured the domain that way—but who also clicked a link or used language like "interested" or "tell me more." Without the secondary-signal requirement, these would have been lumped in with the false replies and discarded. The rule does not say "ignore all catch-all replies." It says "ignore catch-all replies unless a secondary signal exists." That distinction is the entire game.

Start with the mechanism, not the policy. A catch-all domain is a mail server configured with a wildcard MX record that accepts mail for any address at that domain, then routes it to a single inbox or, in the abusive case, an auto-responder that generates a plausible-sounding reply. The DNS lookup is your first and cheapest filter. According to the Stanford AI Lab's 2025 methodology, a simple `dig mx` query or a tool like MXToolbox reveals the wildcard pattern before you ever send a message. If the MX record returns a catch-all flag, you know in advance that any reply you receive is suspect. This is not a post-hoc correction; it is a pre-send classification that changes how you interpret the response stream.

Rule 1 is therefore non-negotiable: perform a DNS lookup on every domain before sending. A custom script that checks for wildcard MX records takes minutes to write and saves hours of downstream scoring cleanup. The cost of skipping this step is not a bounced email—it is a false reply that enters your pipeline as a lead and inflates your scoring model with noise. The high false-reply rate from catch-all domains is not a random artifact; it is a structural property of how these servers behave, and it is entirely predictable from the DNS record.

Rule 3 is a hard kill switch. If a catch-all reply contains a generic phrase like 'unsubscribe', 'remove me', or 'not interested', treat it as a false positive regardless of any other signals. These phrases are unambiguous. They do not require interpretation, and they do not benefit from a secondary-signal check. The presence of a negative keyword overrides everything else. This rule prevents the edge case where a prospect clicks a link in your email out of curiosity but then writes "not interested"—the click is a secondary signal, but the sentiment is negative, and the negative sentiment wins.

How to Choose Well

Rule 5 is the feedback loop. Continuously sample and label your replies to measure your false-positive rate. If the rate for catch-all domains exceeds a certain threshold (roughly a third), your scoring threshold is too lenient or your required secondary signal is too weak. Adjust accordingly. This is not a one-time calibration; it is an ongoing process. The false-reply figure from the Stanford study is a baseline, not a constant. Your domain mix, your email copy, and your audience all shift the rate, and you need to track it monthly to keep your scoring model honest.

The decision tree is short and mechanical. DNS lookup first. Secondary signal second. Negative keyword overrides. High-value manual review as an exception. Continuous monitoring as the governor. Each rule is a gate that filters noise before it reaches your scoring model. The myth that a reply is a reply—that any response indicates interest—is precisely what the high false-reply rate debunks. A catch-all server can generate a reply that looks like a human wrote it, but it cannot generate a click, a positive sentiment, or a direct question. Those signals are the only things you can trust.

Rule 3 is a hard kill switch. If a catch-all reply contains a generic phrase like 'unsubscribe', 'remove me', or 'not interested', treat it as a false positive regardless of any other signals. These phrases are unambiguous. They do not require interpretation, and they do not benefit from a secondary-signal check. The presence of a negative keyword overrides everything else. This rule prevents the edge case where a prospect clicks a link in your email out of curiosity but then writes "not interested"—the click is a secondary signal, but the sentiment is negative, and the negative sentiment wins.

Rule 4 introduces a manual review exception for high-value accounts. For accounts with an ACV above roughly $50k, the cost of a false negative—missing a real lead—is higher than the cost of manually reviewing a catch-all reply that lacks a secondary signal. In these cases, a human should read the reply and make a judgment call. The mechanism here is asymmetric risk: a false positive costs you scoring accuracy, but a false negative on a $50k account costs you revenue. Manual review is expensive, but it is cheaper than losing a deal. This rule is not a loophole; it is a deliberate trade-off between precision and recall, and it should be applied only to the top tier of your account list.

Rule 5 is the feedback loop. Continuously sample and label your replies to measure your false-positive rate. If the rate for catch-all dom

Frequently Asked Questions

According to the 2025 EmailListValidation.com study, what percentage of business domains are catch-all?

11.3% of business domains are catch-all.

In the Stanford AI Lab 2025 study, what was the inflation factor for false replies from catch-all domains relative to non-catch-all domains?

The inflation factor was 9.5x baseline.

What specific secondary engagement signal must be present before a reply from a catch-all domain with an auto-responder counts toward a lead score?

A link click or positive sentiment keyword must be present before the reply counts.

How should a reply from a catch-all domain configured with a spam trap be treated in lead scoring?

Exclude entirely and flag the domain.

What type of messages did the SalesLoft 2024 report find in a substantial portion of catch-all domain replies?

Generic 'unsubscribe' or 'remove me' messages.

What does ORMIndex™ measure and on what scale?

ORMIndex™ measures search, reviews, social sentiment, and digital presence on a 0-100 scale.

Quick answers

Sources: Reddit, Reddit, Forbes, Reddit, Reddit

Also worth reading: How to get a free personal email domain for your custom address: How to get a free · The simple guide to setting up a professional email domain: simple guide to setting up · B2B Email Marketing Benchmarks 2024 Key Metrics and Industry Averages Revealed: B2B Email Marketing Benchmarks 2024