Thanks again, MC team!

However, I do think this was a good, semi-scientific test. Maybe it was just a bad blog title. The actual question from our customer was, “do spam filters penalize me for bad alt-text?” Which is a good question.

Instead of just telling them the seemingly obvious answer (“you should assume that spam filters will just read everything as flat text”), I decided to run an experiment.

I took my own campaign that scored very well (good content, healthy mix of text + images), and added the most disgustingly bad alt-text descriptions to several images. I used words and concepts that gave very high spam scores found at:

http://spamassassin.apache.org/tests_3_0_x.html

Then, I passed it through our inbox inspector, which scanned it in Spam Assassin, Postini, Barracuda, Cloudmark, MessageLabs, and other major spam filters. Not one of them (except, as one user pointed out, the one with the espanol criteria) called me out on my alt-text content, and my score was still _technically_ a passing score.

Doesn’t mean I’m right and you and J.D. are wrong. Or vice versa. Maybe spam filters read alt-text, but then ignore it. Maybe they’ve learned that alt-text is usually ignored by lazy spammers, so the mere presence of alt-text is a good thing to them (even if the content’s bad). Maybe they ignore anything inside an IMG tag. Maybe the inbox inspector was off that day. But if you CTRL+F around that spamassassin criteria page above, there’s no mention of alt-text that I can find (not that the list there is all inclusive).

FTR, my point isn’t that senders can go nuts with alt-text. It just means that as I’m coding my HTML emails, alt-text is one less place I have to worry about triggering something by accident. So, when I paste a photo of my programmers having a water-gun fight in the office, I don’t have to pull out a thesaurus to come up with alternatives for “hardcore nerds having wet fun around the office.” It’s a decent alt-text description that might actually convince recipients to turn images on.

Maybe I can do some more testing. Like removing the alt-text completely, to see if leaving it blank hurts my score worse than putting in nasty stuff.

Thanks for your input, and never hesitate to call me an idiot.
All conversation is good, right?

My understanding of email filters and spam engines (some of this is from actually reading some of the OS engines’ codebase) has always been that they simply read the HTML in the email as flat text.

This would lead me to strongly agree with JD’s post above.

From a technical view point your post seems quite unsavvy in both its original question and its tested result.

Sorry to say the above as i love the service to death. It just worried me that such a technically minded company would post such a non-technical experiment – it leads us to believe you are not the oober dojo ninja’s we once thought you were. And thats bad – as chimp ninja’s are cool…

I have no idea why that’s in Spanish, but it means “Commands/exclamations in all caps”

That’s 35% of your not-so-great SpamAssassin score.

I think what your test actually showed was that those particular bad words aren’t, by themselves, sufficient triggers for spam filters. That makes sense, since most of ‘em stopped doing raw keyword filtering long ago and moved towards other methods.

Humans, however, still read those words when they decide whether to open a message or report it as spam — so a campaign offering “FREE HOT WET VIAGRA!!!!!” will end up with poor delivery.