Do images in cold emails hurt deliverability? Not by themselves. Every vendor who has published data on the question lands in the same place: the identical email with one small, self-hosted image reaches the same folder as the version without it. What actually moves placement is the stuff that tends to arrive alongside images: a message that is mostly picture, a stack of heavy files, an image served from a domain with a dirty reputation, and the invisible tracking pixel most sending tools inject without asking. This article is the images verdict on its own terms. We have already ruled on links in cold emails and on video in cold email, and the answer for images is more permissive than either.
The verdict comes from sending, not from a tool vendor’s marketing blog. Our parent agency, Referral Program Pros, has run more than 4,000 outbound campaigns and booked over 7,000 meetings, and GTM Bud is that agency’s playbook turned into software, backed by a written guarantee: 5 percent positive replies on LinkedIn, 1.5 percent on email, or a full refund. Our own practice across those campaigns is to default to plain-text-looking emails for first touches, a practice statement rather than a statistic, and this article explains why that default survives contact with data that says images are mostly harmless.
Do images in cold emails hurt deliverability?
Images in cold emails do not hurt deliverability on their own. The published vendor tests agree on this: Lemlist reports no deliverability difference between emails with and without images across its own sending data, and Outboundly ran its own placement check and titled the result “No, We Checked.” What filters score is not the presence of a picture but four surrounding variables: the image-to-text ratio of the message, the number and file size of the images, the reputation of the domain hosting them, and whether a hidden tracking pixel rides along. Keep all four clean and one image is a rounding error. Let any of them slip and the image becomes the visible part of a shape filters are trained to catch. One honest caveat belongs next to that verdict: no controlled public dataset isolates images as a deliverability variable in cold email, so every number in this debate is vendor-published, and the verdict rests on their agreement rather than on any single study.
The rest of the article unpacks those four variables, pins the circulating numbers to their real sources, and ends with the rule we actually run.
Where do the image deliverability numbers come from?
Almost everything quotable about images in cold emails was published by companies that sell sending tools. That does not make the findings false, but you should know what each one measured before you redesign your sequence around it.
| Source | What it claims | What to know before quoting it |
|---|---|---|
| Lemlist analysis | No deliverability or open-rate difference with or without images; one image lifts replies by almost 20 percent | Vendor-published and observational, no methodology or controls released; senders who add one good image may simply write better emails |
| Outboundly placement test | No measurable impact from images on inbox placement, per its own check | Vendor-published; a placement spot-check, not a controlled study |
| Smartlead data breakdown | Images alone do not trigger filters; ratio, file size, hosting domain, and tracking pixels do | Vendor-published synthesis; its thresholds inherit from marketing-email filtering practice |
| Repliq, Snov, Autoklose guides | One or two small, relevant images are acceptable in cold campaigns | Vendor knowledge-base guidance, consistent but not data |
| Hunter analysis | In a study of 2.2 million emails, heavier HTML and graphic elements were associated with dramatically higher bounce rates | Correlational; measures HTML weight overall, not images in isolation |
The loudest claim in that table deserves its provenance attached. The near-20 percent reply lift from adding one image comes from Lemlist, a sending tool, analyzing emails sent through its own platform over one month. It is a real published finding and it is not a controlled experiment: nothing separates the effect of the image from the effect of being the kind of sender who uses one well. Quote it as “Lemlist reports,” never as “studies show.”
And state the gap plainly: there is no controlled public dataset that isolates images as a cold email deliverability variable. The honest reading of the evidence is a consistent direction across vendors with magnitudes nobody has verified independently. That direction is still useful, which is what the next section is for.
What actually gets image-heavy cold email filtered?
The four variables the sources agree on form a practical checklist. Each row is a dial, and the danger zone on any one of them can undo clean settings on the others.
| Variable | Safe zone | Where filters start caring |
|---|---|---|
| Image count | Zero or one, two at most, per Lemlist and Snov guidance | Lemlist reports replies dropping at three or more; five reads as a newsletter |
| File size | Compressed, total image weight under 100KB, per Smartlead | Oversized messages trip size-based filtering before content is even scored |
| Image-to-text ratio | Mostly text; Smartlead cites roughly 60/40 text to image as the floor | Past about 40 percent image content, per Smartlead, the message matches the image-spam pattern |
| Hosting domain | Your own domain or a CDN account you control | A hosting domain other senders burned taints every email that embeds its images |
| Tracking pixel | None in cold email | A hidden 1x1 image with a per-recipient URL is the image filters score hardest |
| HTML wrapper | Minimal markup that renders like plain text | A designed template routes mail to Promotions regardless of what the image shows |
Two of those rows are really other articles’ territory, and the boundaries matter. The HTML wrapper row is the format question, and the verdict in our comparison of plain text vs HTML cold email is that the wrapper does more placement damage than any single element inside it. The ratio and size rows are the only rows unique to images, and they share one root: a message that stops looking like a person typing and starts looking like a rendered campaign.
The hosting row is the quiet one nobody checks. An embedded image is a URL the receiving provider fetches and evaluates, exactly like a link, so the image inherits the reputation of whatever domain serves it. Host every image you send on your own domain, and the row never costs you anything.
When is one self-hosted image safe in a cold email?
One image is safe when it is small, self-hosted, surrounded by real text, and sent to a prospect who has already engaged. That last condition is our own line, and it is stricter than the vendor data demands: our agency practice at Referral Program Pros, and the default GTM Bud ships with, is a plain-text-looking first touch with no images at all, because the first email has to earn inbox placement from a stranger and every removable risk should be removed. By the second or third touch, a prospect who opened, clicked, or replied has handed the provider positive signals, and that is where a screenshot of their website, a relevant chart, or a product image can do the persuasion work text cannot.
Note how this differs from the neighboring verdicts. For links, the first-touch ceiling is one and often zero, because a URL is an active request for a click that filters score against phishing patterns. For video, the embed never arrives at all; what lands is a thumbnail image plus a tracking link, which stacks the image question on top of the link question. A static image is the mildest of the three: it asks the recipient for nothing, hides nothing, and carries risk only through the four variables above. That is why the sources are more relaxed about images than about anything else you can put in a cold email.
For agencies running campaigns on client domains, the image rules compound across every mailbox in the pool, which is why cold email for agencies setups should set an image policy once, centrally, instead of per campaign. One client’s 2MB banner in a signature can drag placement for a domain the agency spent weeks warming.
The tracking pixel is the image that matters most
Here is the twist in the images debate: the highest-risk image in most cold emails is one the sender never consciously added. Open tracking works by embedding a hidden 1x1 image with a unique URL per recipient, which means every “image-free” campaign with tracking enabled is actually an image campaign, and the image it contains is the exact shape filters associate with surveillance and bulk mail. Even Lemlist, arguing that visible images are harmless, concedes in the same piece that pixel trackers can increase the chance of landing in spam.
So the most valuable image decision in cold email is turning one off. We took that question apart in our verdict on open tracking in cold email: the pixel costs placement and consent to measure a number privacy proxies broke years ago. Disable it, measure replies, and your campaign’s image risk usually drops to zero before you have discussed a single visible picture.
What transfers from marketing email research, and what does not
Most image-filtering rules were born in opted-in marketing email, and that lineage should shape how hard you lean on them. The image-to-text ratio thresholds, the file size limits, and the image-spam pattern all come from providers filtering newsletters and promotions at massive scale, where senders have permission, engagement history, and dedicated infrastructure. Cold email has none of those cushions. The ordering almost certainly transfers: a mostly-image message is worse than a mostly-text one in any inbox, and a flagged hosting domain hurts everyone. The magnitudes do not transfer: a ratio that is merely suboptimal for a newsletter with an engaged list can be disqualifying for a cold sender whose domain has no reputation buffer, so treat every marketing-email threshold as a ceiling, not a target. Google’s sender guidelines apply to both worlds and set the floor either way: format messages to standard, do not hide or obscure content, and make the message match what the recipient can see.
The practical consequence is the asymmetry we keep coming back to. The upside of an image in a cold email is a possible lift one vendor measured on its own traffic. The downside, if any of the four variables slips, is the inbox placement every other metric depends on. The infrastructure that placement sits on, authentication, warmup, and volume discipline, is its own project, covered in our cold email deliverability guide; the image decision is just one input riding on top of it.
Frequently asked questions about images in cold email
Do GIFs hurt cold email deliverability?
A GIF is scored like any other image, so the deliverability math is the same: one small, self-hosted file is tolerable, and a heavy animated loop pushes message weight and campaign signals in the wrong direction. The performance case is weaker than for static images, though. Lemlist, in the same vendor-published analysis that reports a reply lift from one image, reports that emails with GIFs got 16 percent fewer opens. An animation that loops like an ad reads like an ad; if a picture earns its place, make it a static, compressed screenshot.
Where should you host images used in cold emails?
On a domain you control, or a reputable CDN tied to your own account, and nowhere else. The receiving provider fetches every embedded image from its hosting URL and scores that domain the way it scores a link, so an image served from a domain other senders have burned taints your email through no fault of your copy. Smartlead names hosting-domain reputation as one of the variables that actually moves placement. Never hotlink from third-party sites, and never rely on a sending tool’s shared asset domain for cold volume.
Can you send a cold email that is only an image?
No. A message that is mostly or entirely picture with little real text is the classic image-spam pattern: spammers historically rendered their pitch inside an image so text-based filters could not read it, and filters have been trained against that shape for two decades. Smartlead’s breakdown flags the image-heavy ratio as the point where providers start paying close attention, and Google’s sender guidelines tell senders not to hide or obscure content. Write the message in text and let an image illustrate it, never replace it.
Do images affect cold email open rates?
Lemlist’s vendor-published analysis of its own sending data reports no difference in open rates between emails with and without images, which matches its deliverability finding. The deeper issue is that open rates are no longer a trustworthy instrument at all, because privacy proxies fire tracking pixels whether or not a human read the message. Judge image decisions by reply rate per touch, the number that cannot be faked by a proxy server.
Should images go in the first cold email or in follow-ups?
In follow-ups, if anywhere. The first touch has to earn primary-inbox placement from a stranger with zero engagement history, so it should look like a short plain note from one person to another. By mid-sequence, a prospect who has engaged gives providers positive signals that absorb the extra weight of an image, and Lemlist’s own guidance is to spread images across the sequence rather than stack them in one send. If you would rather have someone enforce that sequencing per campaign, a done-for-you outbound service makes those calls for you.
Does removing all images guarantee inbox placement?
No. Images are one modest input among many, and a plain-text email from an unauthenticated domain blasting a bought list will be filtered with zero images attached. Placement is decided first by authentication, domain reputation, volume discipline, and targeting, and only then adjusted by content signals like images and links. Strip the images by all means, but spend the saved attention on the fundamentals that actually carry the campaign.
Keep the first touch plain, earn the right to show pictures
So, do images hurt cold email deliverability? One compressed, self-hosted image in a mostly-text email does not, and every vendor who has checked agrees, even if nobody has run the controlled study that would settle the magnitudes. The real image rules are about everything around the picture: stay far below the image-heavy ratio, keep total weight small, host on your own domain, and kill the tracking pixel that most tools smuggle in as a hidden image. Then apply the sequencing discipline the data cannot give you: a plain first touch, pictures only after engagement, and replies as the only score that counts.
That is the default GTM Bud’s cold email automation tool runs out of the box: plain-text-looking first touches, no tracking pixel, and copy built to earn a reply rather than decorate an inbox. It is the same playbook our agency used across 4,000+ campaigns and 7,000+ booked meetings, under the same written guarantee: 5 percent positive replies on LinkedIn, 1.5 percent on email, or a full refund. Send fewer pictures at strangers, land more mail in front of humans, and save the visuals for the deck you share after they answer.