Google Is Blocking Sentences, Not Emails

I was on a live coaching call recently and a guy walked me through something that should scare every cold emailer running campaigns right now.

His warmup inbox scores were sitting at 99 and 100 across the board. His domains were aged over a year. His leads were freshly cleaned. He had a standby domain he'd been warming up in the background specifically for emergencies. By every conventional deliverability metric, this guy was doing everything right.

And his open rate had cratered to 4%.

I asked him the obvious questions. Are you on any blacklists? He didn't know. Are your scripts getting blocked? He hadn't thought about that. He kept pointing to the warmup scores as evidence that nothing could be wrong. That's the trap. That's exactly what I want to talk about.

Warmup Tools Don't Track What You Think They Track

Here's what warmup tools actually do: they send fake emails back and forth between inboxes they control, then measure whether those fake emails land in the primary inbox. That's it. That's the whole mechanism.

What they don't track is whether a specific piece of text inside your actual cold email script is getting flagged by Google's filters.

Google isn't just looking at your domain or your IP anymore. It's reading the content of emails. And what we've seen recently - and this showed up clearly with this guy on the call - is that Google will identify a specific sentence, a specific phrase, or even a specific URL inside your email, hash it, and then silently dump any message containing that string straight into the void. Not spam. Not promotions. Gone.

The warmup emails? They have completely different text. So they sail right through. Your warmup dashboard shows green. Your deliverability looks clean. Meanwhile, your actual outbound emails with your actual script are getting killed before they ever reach a human being, and you have no idea.

He had a 4% open rate. His warmup was at 99%. Those two things make no sense together unless you understand this mechanism. Once you do, they make perfect sense.

The Specific Error That Cracked the Case

While we were troubleshooting, he mentioned something he'd almost glossed over. There were a handful of bounce errors he'd seen in his sending tool that said his URL was being blocked. Just a few. Then they disappeared and he couldn't find them again.

That was the tell.

He was sending from multiple domains - which is correct, that's what you're supposed to do to protect volume and isolate risk. But every single one of those emails contained a link back to his main domain. So even though the sending domains were different, every email was effectively tied together through that one URL. Google saw thousands of emails from different senders all pointing to the same place, and flagged the destination domain. Or possibly the URL string itself.

All of his sending domains were sharing one poisoned link. The diversification he'd built was completely bypassed.

Take out the link. Test the same script with no domain reference. See if open rates come back. That's the diagnostic. If they do, you've confirmed the URL is what's being hashed and blocked. Then you either create clean forwarding domains that don't trace back to your main site, or you restructure the call to action entirely so the link isn't in the body of the email.

This is also why I tell people: if you're linking to a domain in cold email, use a domain that looks like it's a real separate thing. Not a bit.ly shortener - those are tracked and flagged even faster. Something that looks like a legitimate standalone domain. The goal is that nothing in your email, not the sender domain, not the linked domain, not the text itself, creates a pattern that Google can cluster and kill.

The SORBS Problem and What It Means for Your Domains

We ran his sending domain through a mail testing tool on the call - mailtester.com will show you whether you're on major blacklists. One of his domains came back listed on SORBS.

SORBS is one of the older, more basic blacklists, but being on it matters because Microsoft and Outlook accounts actively reference it. If you're on SORBS, a significant chunk of your list - anyone using Outlook-based email - is likely rejecting your emails at the server level before delivery even happens. That's not a spam folder issue. That's a hard block. The email doesn't land anywhere. It bounces, or it gets silently discarded.

He'd been attributing some of his bounces to lead quality. Understandable mistake. But what was actually happening is that some of those bounces were the recipient server recognizing his domain, checking it against a blacklist, and bouncing the message back pre-delivery. The lead wasn't bad. The domain was flagged.

Once a domain is blacklisted like that, the math changes. You're not going to rehabilitate it quickly. The move is to spin up new domains, warm them up properly, and move on. He already had a standby domain he'd been warming in the background - smart setup. That becomes the new primary while the others get retired or slowly rehabbed.

If you haven't checked your sending domains against a blacklist tool recently, stop reading and do it now. Mailtester.com or MXToolbox will both show you where you stand.

Spintax Isn't Optional Anymore

I had another person on the same call - different situation, but related problem. Open rates were running 60-65% for weeks. Solid. Consistent across a five-part sequence, consistent across twelve domains. Then in one week, open rates dropped by almost half.

Nothing had changed in the script. Same subject line. Same opening line. Same sequence.

That's the signature of content-level blocking. When your numbers drop across all domains simultaneously and nothing else changed, it's not a domain reputation issue - individual domains fail independently. When everything drops together, something consistent across all of them is getting flagged. And the one thing that was consistent across all twelve domains was the script.

His spintax was minimal. He'd swapped out a word here and there. An opening greeting, maybe a phrase in the middle. That's not enough.

Google's filters are identifying the pattern. Even if individual words change, if the sentence structure is identical, the semantic fingerprint is the same. The fix is aggressive spintax - not just swapping "Hi" for "Hello," but spinning entire phrases, restructuring sentences, varying the flow so that each email is genuinely different from the last.

The fastest way to implement this is ChatGPT. Paste your full email sequence into the prompt and tell it to add spintax at a minimum 90% variation rate, and explain what spintax format your tool uses. It'll rewrite the whole thing with proper bracket syntax in a few seconds. What would take you an hour of manual work takes thirty seconds. There's no reason not to do it.

For that guy, the fix was: add heavy spintax immediately, turn off any domain whose warmup score had dropped below 90 (and his were dropping, a few had dipped into the 80s), and spin up fresh domains to replace them. Anything under 90 on warmup score means you're likely getting manually marked as spam, and it's fighting the warmup system. Once that starts, the domain is in a losing battle. Cut it. Replace it.

Your warmup scores should be 99 or 100. That's not a high bar. That's the floor. If you're at 95 and thinking "that's pretty good," you're wrong. Something's wrong. Find out what before you keep sending.

The Tool Problem Nobody Talks About

One of the guys on the call was using Woodpecker as his sending platform. I had to tell him directly: Woodpecker has had major deliverability issues ever since they lost access to the Google API. That's not a knock on the team, it's a structural problem. The same thing happened to Lemlist - they actually spun their warmup feature out into a separate product specifically because Google was going to ban the integration if they kept bundling it.

This matters because the tool you use to send is part of your deliverability stack. It's not neutral infrastructure. If your sending tool has been flagged or has known issues with major email providers, every campaign you run through it inherits that problem.

Instantly and Smartlead both have deliverability monitoring built in. They'll show you blacklist status, spam folder rates, and flag when something's going wrong. That's the baseline now. If your tool doesn't give you that visibility, you're flying blind by choice.

I'm building this into ScraperCity as well - the infrastructure side of this is something every serious sender needs baked into their workflow, not bolted on after something breaks.

Domain Extensions and the Details That Actually Matter

Quick one because it came up on the call and I see people get this wrong constantly.

Use .com. That's it. Not .io, not .ai, not .co. Dot com has the highest deliverability by a significant margin. If you're spinning up sending domains and you're buying .io because it's cheaper or you think it looks more modern, you are trading deliverability for aesthetics. Don't do it.

Dashes in domains are fine. Numbers are fine - our old agency domain was X27 Marketing and it had the number right in the name, never caused a problem. The rule is just: don't make it look spammy. If a human being would look at your domain and think "this looks sketchy," Google probably agrees.

What you're trying to do is create domains that look like they could plausibly be real companies. The more legitimate it looks, the less friction you face from the filters before a single email goes out.

What "Unsubscribe" Actually Means in Cold Email

One more thing from the call that's worth addressing because I see people confused about this.

Someone asked whether they needed an unsubscribe link in the email body, or whether "reply UNSUBSCRIBE" was sufficient.

My interpretation of CAN-SPAM is that you need an easy, functional way for someone to opt out. A reply-based opt-out satisfies that requirement. You don't need a formal link in the body.

More importantly: if you're writing cold emails the right way, they shouldn't look like mass emails at all. They should read like you sat down and wrote this directly to one person. The moment you add an unsubscribe footer that looks like a marketing email, you've broken that illusion. You've told the reader - and the spam filter - that this is a bulk send.

If you feel like you need a formal unsubscribe link for compliance reasons specific to your market, put it in the signature at the smallest possible font. Don't make it a feature of the email. Make it invisible unless someone's specifically looking for it.

And if you're sending into markets with stricter regulations than CAN-SPAM, that's a separate conversation. Get actual legal advice for your specific situation. I'm not your lawyer.

How to Diagnose This in Your Own Campaigns

If you're reading this and wondering whether any of this is affecting your campaigns right now, here's the sequence I'd run:

First, check your warmup scores. Not just "most are good." Every single sending domain should be at 99 or 100. Run this check today.

Second, run your domains through a blacklist checker. Mailtester.com. MXToolbox. If any domain comes back blacklisted, that domain is likely responsible for a chunk of your missing opens and your unexplained bounces.

Third, look at your open rates across domains. If the drop is inconsistent - some domains fine, some not - it's a domain-specific problem. If everything dropped together at the same time, your script is probably what's getting blocked.

Fourth, test by elimination. Take your current script and strip out every link. Send from a fresh domain to the same leads. If open rates recover, the URL was the problem. If they don't recover, something in the text itself is triggering filters. Start spinning out different versions of each paragraph to isolate which section is the issue.

Fifth, if you're confirmed on a blacklist - spin up new domains. Warm them up. Don't try to rehab a poisoned domain while simultaneously trying to run active campaigns. You'll burn leads and time.

The reason most people never figure this out is that they're looking at the wrong data. They check warmup. Warmup is fine. They assume the problem is leads or subject lines or the offer. They rewrite the pitch. They rebuild the sequence. They get new leads. Open rates stay flat. Because the script is still blocked and they've never isolated it.

The warmup score being green tells you exactly one thing: the warmup emails are going to the primary inbox. It tells you nothing about your actual outbound script. These are two completely different things that happen to be reported by the same tool, which is why everyone conflates them.

Fix the script first. Then fix the domains. Then - and only then - worry about the offer.

If you want to go deeper on the lead side of this, the Best Lead Strategy Guide breaks down how to build lists that hold up over long campaigns, which matters a lot once your deliverability is clean and you need volume that doesn't degrade. And if you're starting from scratch on outbound, the top cold email scripts give you templates that are built to avoid a lot of these trigger patterns from the start.

For serious scaling - the infrastructure conversations, the campaign diagnostics, the offer work - that's what we do inside Galadon Gold. Live calls, real campaigns, real problems. Not theory.

Now go check your warmup scores.