Nearly 25% of corporate press releases disseminated in 2024 were AI-generated, according to a Stanford University study published in October 2025. However, even though these curated announcements were carefully crafted by communications professionals, many never even reached their intended audience. Why not? Simple – spam filters can sniff out patterns in AI-generated text, and they're automatically designating press releases as junk mail.

And this isn't a theoretical problem. Companies pay thousands of dollars to distribute a press release – only to watch their announcement vanish into a void, damaging relationships with media and brand visibility, all without any indication of what went wrong.

The AI Content Detection Arms Race

Today's email security platforms are shockingly adept at detecting AI-generated text. Gmail recently rolled out an entirely new spam filter which leverages a proprietary text classification stack called Resilient & Efficient Text Vectorizer. One of its purposes? Detecting adversarial text manipulations. But this is a double-edged sword, as research published in March 2025 found that while AI-generated fishing emails bypass this filter 86% of the time, regular generated content is more likely to be caught in the crossfire.

It's a cruel irony. Spam filters rely on natural language processing to analyze sentence structure, tone, and context. At the same time, communicators have been using AI tools like ChatGPT to generate press releases – creating written content that follows the same patterns spam filters are now programmed to detect. And instead of getting in front of potential investors, executives, and journalists who would be interested in a business' news, a company's carefully curated release gets flagged as an attempted fish.

Consider some of the kludgy linguistic tells security spam filters are programmed to look for: the use of imperative verbs to create artificial urgency, clause density that remains too consistent, pronoun usage that follows unusual patterns, sentence and paragraph structures that repeat inside the same piece, etc. And what do many AI writing tools currently produce by default? Stilted, robotic text that follows non-standard patterns. In fact, a recent study using XGBoost was able to achieve 96% accuracy at detecting AI-generated content by taking stylometric measurements of 1,200 English NLM samples.

The Hidden Cost of Efficiency

Using AI press release generators can seem like a good way to save time. These tools claim to cut writing time from hours to minutes. They help keep your company's voice the same across releases. They also promise to give your team more time for important work. But this efficiency has risks you need to think about.

Most people don't know the deeper problems when releases go to spam. Email providers use metrics such as open rates, replies, and spam clicks, to keep track of sender reputation. Every time your release is marked as spam, your company's domain gets a lower sender score. This creates more filtering, and not just for press releases – it can also affect replies and even important business emails.

The effect builds over time. Industry research estimates spam costs $20.5 billion each year from lost productivity and technical costs. For companies that reach out by email to journalists, investors, partners and more, a poor sender score can make entire email channels useless.

AI-generated fishing attempts during recent testing, blocked 9 out of 10 emails from Yahoo's spam filter, and at the same time, 58–66% of real emails generated by AI were marked as spam. With over 50% false positives, Yahoo's stance is clear: if in doubt, block it.

Why AI Content Triggers Spam Filters

Today's spam filters use more than just keywords. They use machine learning to look at billions of emails. They also look at sender behavior and links. Apps like Gmail and Outlook, and enterprise security filters like Barracuda, use behavioral analysis, anomaly detection, and deep learning to find threats.

AI-generated press releases have signs these tools can pick up on. The language is often too polished. There's not enough variety in the writing. The tone stays the same. The minor errors that come up when people write are missing.

Studies show that current AI detection tools look for features like the frequency of transition phrases, overall punctuation, comma distribution, variation in sentence length, use of passive or active voice, word choice, and paragraph structure. When these factors fit together too well, filters take note.

There's another problem with the amount. Companies that use AI can suddenly start putting out tons of press releases. Spam filters look for big increases in content publication, because this matches spammer behavior. What feels like new efficiency to your comms team looks like a mass spam campaign to security filters.

Protecting Your Communication Channels

You don't have to stop using AI tools. The key is to use them carefully and still sound real. Before you send any press release made by AI, use an AI Humanizer ( It can change the way you write, add some variety, and help you avoid filters.

You also need to keep a human in the loop. Your press release from AI should be a first draft, not what you send out. Add details that your industry or company uses, write quotes that sound like you, and change how you write. When you can, use quotes from real people at your company instead of made-up statements from AI. These steps show filters and people that the pitch is real.

You should also check your technical authentication. Use your own domain to send out press releases. Make sure your domain has the right tools, like SPF, DKIM, and DMARC records set up. These help email providers know your identity and make it more likely that your press release goes through. Check your sender reputation with tools that show what email providers think of your domain.

Don't make big changes fast. If you usually send out one press release a month, don't suddenly start sending out 10 every week. Go slow. Try not to send a lot of emails in a row. Wait and only send news releases when you have something new to say. This matches the ways you are used to making new content. It also matches the way search engines look at new releases and how email providers check if you are real.

To keep your engagement, update your media lists. Get rid of emails that are no longer working for your company and keep track of bounces. Focus your pitches on people who actually open your press releases. High open rates help keep your pitch from being flagged.

The Path Forward

Things are changing fast. Soon, some experts say that 25% of people searching for info will stop using Google search. They'll start using new“answer engines” like Perplexity and ChatGPT. We're getting to a place where we need a whole new strategy for press releases. They have to make sense to both the people who read them and the bots that put your content back together.

It's a new way to look at press releases. Instead of writing only for journalists or Google, PR pros need to figure out how to make releases that work for these AI engines. That means writing out clear and structured info and making sure your data has the right schema.

The basic idea is: If you only care about being quick, you'll destroy your ability to be heard. These AI tools are great helpers, but they can't replace your judgment.

Companies that see that will keep getting through. Companies that use them to put out more junk than humans ever could will see their reputation die a slow, silent death-one blocked press release at a time.

Frequently Asked Questions

Can I use AI at all to write my press releases?

Yes! But use it carefully. Write a first draft with AI and then have a human change it a lot. More than anything, add a real-sounding voice, change up the way you write, and bring in problems and mistakes that only a person would write. Putting your press release through a humanization tool ( ) before you send it out will help keep you off the spam list.

How do I know if my press releases are being blocked or put in spam folders?

Check your sender reputation score with email deliverability tools, look at your bounce rates and open scores, and send tests to different emails to see if they go through. Low open rates with no bounces often mean you're going to spam folders. Also, try to get a few friendly reporters to check their spam folder right after you send your press release.

MENAFN31102025003238003268ID1110278218