How to Audit the Email Automations That Are Still Sending

Published July 27, 2026

Part of: Email Marketing — our full guide on this topic.

Disclosure: Some links below are affiliate links. If you sign up through them we may earn a commission at no extra cost to you. We only recommend tools we'd genuinely suggest to a friend. See our full disclosure.

You set up a welcome sequence. It worked, so you left it alone. That was the point — email automation is the part of the business you build once and stop thinking about.

Now think about it for an hour.

Somewhere in that sequence there is probably a link to a product you retired, a promise to send something you no longer make, a line saying you publish every Tuesday, or a sign-up form belonging to an email tool you left. It has been sending in that state to every new subscriber since the day it broke, and nobody has told you, because the people receiving it have no idea what it was supposed to say.

This site has a whole playbook for the subscriber who goes stale. It has nothing about the email that goes stale. Here is that pass.

The quick version

• A sent email cannot be edited. It is the only asset you own that you cannot fix retroactively — which changes what “fixing it” even means.
• Your automations are the only thing you own that goes out and finds people. A stale page waits to be found; a stale sequence arrives uninvited at the best moment you get.
• The only honest inventory is to subscribe to yourself from a fresh address, through the real form.
• Check both directions. Half the faults are emails still sending that shouldn’t be. The other half are emails that should be sending and aren’t — and those are silent.
• Read them in the order they reach people, not in the order they are listed in your tool.
• Editing a live sequence is not free — people are inside it right now, and what happens to them depends on your platform.
• Point sequences at pages you control, not at things. That is the fix that stops this recurring.

Why this is the one thing you cannot fix retroactively

Everything else this site tells you to maintain is a live document. An old blog post can be updated, and the update is what the next visitor sees. A dormant social profile can have its bio rewritten and the rewrite is instantly the truth. Even a retired product can keep a working URL that explains itself.

An email is a snapshot. Once it has sent, it is a copy sitting in somebody’s inbox, and there is no version of “fixing it” that reaches them. You can correct the sequence for everyone who comes next, and that is genuinely worth doing, but the people who already received the broken version received it permanently.

That has one practical consequence and it is the reason this audit is worth an hour rather than a shrug: the only variable you control is how long the broken version keeps sending. Not whether it sent — that is already decided. Just the length of the interval between the day it broke and the day you look. Every month you don’t look is another cohort of new subscribers whose first impression of you was an email pointing at a 404.

It also explains why the prevention section at the end of this article matters more here than in any of the other maintenance passes on this site. Elsewhere, prevention saves you effort. Here, it saves you the emails you would otherwise have already sent.

What “cleaning your email” usually means — and what it leaves out

Search this site for advice about tidying up your email and you will find a lot of it. All of it is about the people.

Win back inactive subscribers is about the subscriber who stopped opening. The advice on list hygiene is about removing dead weight so your deliverability holds up. Segmentation is about sorting people by what they have done. Open rates measure how people respond.

Every one of those treats your list as the thing that decays and your emails as the fixed reference point. The reverse case is at least as common and gets no attention at all: the list is fine and the email is wrong.

(There is a third case, outside this article’s scope but worth knowing exists: the subscriber who asks you to stop holding their details rather than to stop receiving them. That one is handled differently again, and an automation is one of the places their data hides.)

There is a reason for the imbalance, and it is worth naming because it explains why this goes unnoticed for years. A subscriber who goes cold generates a signal — the open rate falls, the number sits there in a dashboard, eventually you notice. An automation that goes out of date generates nothing. It keeps sending. Delivery is fine. Nobody replies to say “that link is dead,” because a new subscriber does not know what was supposed to be there. The absence of complaints reads exactly like everything working.

So the audit cannot be triggered by a metric. It has to be scheduled, or triggered by an event you already know about.

The only honest inventory: subscribe to yourself

If you have run the software subscription audit, you started from twelve months of bank statements, because money leaves a trail. If you have run the dormant profile pass, you started from searching your own name, because free accounts leave no financial trail but do leave a public one.

An automation leaves a trail in exactly one place: somebody else’s inbox. So put yourself in it.

Take an email address that is not already on your list — a spare account, or an alias if your provider supports them. Then go to your own site, in a private browsing window, and sign up through the real form the way a stranger would. Not through a test link. Not by adding yourself in the back end. Through the form, on the live page.

Then leave it alone for a week and read what turns up.

This is the whole method, and it is better than anything you can do from inside your email tool for a specific reason: it tests the chain, not the parts. Opening the sequence in the editor shows you the emails. Subscribing shows you whether the form still applies the tag, whether the tag still fires the trigger, whether the thing you promised actually arrives, how long the gaps really are, what name it comes from, what the reply-to does, and how it renders in a real inbox rather than a preview pane. Any one of those links in the chain can be the broken one, and only two of them are visible in the editor.

Two notes on doing it properly:

• Do it once per entry point. A lead magnet form, a newsletter footer, a checkout, a free email course landing page and a link-in-bio form may all feed different automations, and the one you never think about is the one most likely to be broken. Make a list of every place someone can join your list — that list alone is usually a surprise.
• Resist the urge to fix things while they arrive. Read the whole sequence first and take notes. Editing halfway through changes what the rest of your test receives, and you lose the one clean run you had.

What actually goes out of date

Six things, roughly in order of how much they cost you.

1. Links to things that no longer exist. The big one, and the one with a direct revenue consequence. Every sequence you have written contains links, and links point at things that move. A product you retired. A membership you closed, whose welcome and onboarding sequences are still cheerfully greeting people into something that no longer exists. An article you merged or deleted. A sign-up form on the email platform you switched away from. A file link to a lead magnet you rebuilt. Click every one of them, from the received email, not from the editor.

2. The offer. A sequence written to sell a thing is a sequence about that thing. If you now sell something different, the sequence is still doing its job perfectly for a product nobody can buy — and it is doing it at the moment a new subscriber is most likely to buy something. That is not a broken link, it is a whole email spent on the wrong thing.

3. Facts about you with a shelf life. “I publish every Tuesday.” “I’m running a cohort in the spring.” “I’ve been doing this for three years.” “Reply and I’ll send you my template.” Each of those was true once. The last one is the worst kind, because it is a promise: someone replies, nothing comes back, and you have taught a brand-new subscriber that replying to you does nothing.

4. Prices and numbers written into the email. A price in an email is frozen at the moment it sent. If your pricing has changed, everyone receiving that sequence is being quoted a number you will not honour, and finding out at the checkout. Prices belong on the page, and the email should link to it. Sequences are the single most likely place a dead price survives a change, which is why the price-rise checklist treats them as a mandatory stop.

5. Sender identity. The from-name, the reply-to address, and the signature. These are set once, usually at account setup, and never looked at again. Check that the reply-to is an inbox you actually read — a reply that lands somewhere unmonitored is worse than no reply prompt at all — and that any links in your signature still resolve, since signatures commonly point at a social profile or a bio page that has since moved.

6. Collisions between automations. Two sequences firing on the same person is the fault you cannot see in either one individually. The classic version: your lead magnet delivery and your welcome sequence both trigger on the same sign-up, so a new subscriber gets two emails within a minute. The other version: someone joins during a launch and receives launch email three before welcome email one. Subscribing to yourself is what surfaces these, because it is the only view that shows the combined stream a real person sees.

The half nobody checks: the automations that stopped sending

Everything above is about emails that send when they shouldn’t say what they say. The other half of this audit is the opposite, and it is genuinely invisible.

An automation that has stopped firing produces silence, and silence is what success looks like. (This is also why automations are the first thing to go missing after any forced move — nothing announces that a sequence failed to come across, which is one of the traps in rebuilding on a platform’s closing deadline.) Nothing arrives in your inbox. Nothing appears in a report. No one complains, because nobody misses an email they never knew was coming. A sequence can sit dead for a year while you continue to believe every new subscriber is being welcomed.

The common causes, all mundane:

• It was paused and never switched back on. Usually while you were editing it, or during a launch, or while migrating.
• The trigger points at something that no longer exists — a tag you renamed, a form you deleted and rebuilt, a segment whose rule no longer matches anyone.
• The form stopped applying the tag. You rebuilt an opt-in form, or embedded a new one, and the new form collects addresses perfectly while doing nothing else. Subscribers arrive and enter no sequence at all.
• It did not survive a tool change. When you move email platforms, the list is what everyone checks, because the list is the asset. Automations frequently do not transfer, or transfer as unpublished drafts. It is entirely possible to run for months on a new platform with a perfectly migrated list and no working automations behind it.

The check is the same one: your test subscription either produces the emails or it does not. If nothing arrives, you have found something more valuable than a dead link. And it is worth doing this test on every entry point precisely because this failure is per-form, not per-account.

Read them in the order they reach people

When you open your email tool, the automations are listed by name, or by date created, or alphabetically. None of those is the order to work in.

Work in order of who is receiving it right now.

1. Automations firing today, for new subscribers. Your welcome sequence and your lead magnet delivery. Everyone who joins this week gets these, and they are the first thing anyone experiences of you. Fix these before you look at anything else.
2. Automations on regular but less frequent triggers. Post-purchase sequences, nurture sequences, anything tied to a segment people enter occasionally.
3. Automations on rare triggers. Your launch sequence, a win-back campaign, anything seasonal. These are the ones most likely to be broken, because most time has passed since they were written — but nobody is being harmed by them today, and a launch sequence should be reread before its launch anyway.
4. Paused and dormant automations, last. They are not doing anything, which makes them safe, which makes them the lowest priority.

This ordering matters because audits get abandoned partway through, and the point of an ordering is to make the abandoned version still useful. If you get through nothing but item one, you have fixed the emails that every new subscriber will receive this month.

Editing a sequence while people are inside it

This is the part with no equivalent anywhere else on the site, and it is where an audit can do real damage if you charge in.

A live automation has people standing in it. Someone subscribed yesterday and is between email two and email three right now. When you change the sequence underneath them, something happens to them — and exactly what happens depends on your platform.

The questions to answer before you edit, not after:

• If I add an email in the middle, do people already enrolled receive it? Platforms genuinely differ here, and some differ depending on whether the new step is before or after a given subscriber’s current position.
• If I delete an email, what happens to someone waiting on it? Do they move to the next one, or does the delay they were waiting through get discarded?
• If I switch the automation off, does everyone inside it stop, or do they finish what they started?
• If I edit the text of an email nobody has reached yet, does the change apply to them? This one is usually yes, and it is why text edits are much safer than structural ones.

Find your platform’s documentation on this, or test it on a spare address you have deliberately parked mid-sequence. Do not guess, and do not assume it works the way your last tool did.

The practical rule that falls out of this: edit the words freely, restructure carefully. Fixing a dead link, correcting a price reference, rewriting a stale sentence — those are text changes inside an email that has not sent yet, and they are almost always safe. Adding, removing or reordering steps is a structural change with real consequences for people mid-flight. If a sequence needs restructuring, it is often cleaner to build the new one alongside, point the trigger at it, and let the old one drain — leaving it running for anyone already inside, then switching it off once it is empty.

That drain pattern is not specific to email. It is the general answer to changing anything while people are partway through it — a course, a cohort, an onboarding flow. Stop new people entering the old version and let it empty on its own, rather than lifting anyone out mid-way, because their position in the sequence is exactly the thing a migration loses.

And whatever you change, run the subscribe-to-yourself test again afterwards. The edit is the most likely moment for a working automation to become a paused one.

Writing sequences that don’t rot

The sequences that survived the audit best are not the ones that were written most carefully. They are the ones that were written to refer to fewer things.

• Link to pages, not to things. A link to your start-here page, your current-offers page or a topic hub keeps working when the underlying product changes, because you update the page once and every email that ever pointed there is silently corrected. A link to a specific checkout, a specific file or a specific dated event breaks permanently the moment that thing moves. This is the same rule as where a bio link should point, for the same reason, and it matters more in an email because you cannot go back and change the copy someone already has.
• Write what you generally do, not what you are currently doing. “I write about building a small online business” survives indefinitely. “This spring I’m running a cohort” is wrong by summer. If a sequence needs to mention something current, put the current thing on a page and link to the page.
• Never make a promise the sequence cannot keep by itself. “Reply and I’ll send you X” depends on you, forever, for every subscriber. If you want to give them X, automate giving them X.
• Keep evergreen and time-bound emails in separate sequences. Mixing them means the whole sequence inherits the shelf life of its shortest-lived email.
• Write the review trigger down when you build it. Not a date — a condition. “Reread this if I retire the toolkit, change email tools, or move the resources page.” That is a line in your operating notes and it is what turns this from an annual chore into a thing that gets caught on the day it breaks. The same notes are the right home for a dated log of what you change across the whole business, which pays off somewhere else entirely: when traffic drops, the most likely cause is something you did a few weeks earlier and have completely stopped thinking about.
• Keep the count low. Every automation is a small permanent maintenance obligation. Five sequences you reread is better than twelve you don’t.
• Know that they outlive the business, not just the offer. The four trigger events above all assume you carry on afterwards. There is a fifth, and it is the one where an unread sequence does the most damage: if you ever stop entirely, every automation keeps introducing a business that no longer exists and pointing at checkouts that no longer work. Switching them off belongs before cancelling the email tool, which is why it sits where it does in the closing-down sequence — a cancelled tool sometimes keeps sending and always stops being editable.

Five mistakes

1. Auditing from inside the editor. Reading your own sequence in the tool that built it shows you the text and hides the trigger, the timing, the delivery and half the links. It feels like the audit and it is the preview.
2. Only checking the emails that send. The dead automation is more expensive than the broken one and produces no evidence at all. If your test subscription is quiet, that is the finding.
3. Restructuring a live sequence before checking what happens to the people in it. Adding an email in the middle can mean a hundred people mid-flight either all get it or none do, and you will not know which until it has already happened.
4. Fixing things as you read them. You have one clean run through the sequence as a real subscriber sees it. Spend it on reading. Take notes, then fix everything in one pass, then test again.
5. Treating this as a yearly calendar item only. Sequences do not decay gradually — they break on the day you retire a product or change tools. The calendar pass catches what you missed; the event trigger is what actually keeps them correct.

The pass, in seven steps

1. List every entry point — every form, checkout, landing page and link that can add someone to your list.
2. Subscribe to yourself through each one, from a fresh address, on the live site.
3. Read a week of arrivals without touching anything. Note dead links, stale facts, wrong offers, bad timing, collisions — and note the entry points that produced nothing.
4. Sort the findings by who is receiving them today, and start with the welcome sequence.
5. Make the text fixes first — links, prices, promises, sender details. These are safe and they are most of the value.
6. Handle structural changes deliberately, having checked what your platform does to people mid-sequence, and preferring build-alongside-and-drain over surgery on a live flow.
7. Retest by subscribing again, because the most common way to break a working automation is to edit it.

An hour, once a year, plus a reread on the day you change something the sequence talks about. It pairs naturally with the other two maintenance passes on this site — the software you pay for and the profiles that represent you — and the three of them use three different instruments: a bank statement for what quietly takes money, a search engine for what quietly speaks for you, and your own inbox for what quietly speaks to people on your behalf.

That last one is the only one that goes out and finds them.

There is a fourth pass that pairs with all three and uses no instrument at all, only a question: where is the only copy? The map you write of your sequences during this audit — trigger, order, delays, purpose — is the single most useful thing to keep outside the tool, because automation logic is the one asset that almost never survives a move in a form you can re-import.