In January 2025 Irene Breen was notified that Instagram was removing the adverts for her company’s pregnancy pillows. On a later call with Meta she was told the problem was the word “pregnancy”.
What it cost to fix
“We had to scrub the word from our entire website… from the product name,” Breen told the BBC. “We had to remove all our URLs which had pregnancy in them and rename them all.” The restrictions were eventually lifted, but only after ten months of conversations with Meta — by which time the business had taken a significant financial hit.
Meta says businesses can run adverts promoting wellness, sexual health and reproductive products, but that restrictions exist to avoid users across different countries and cultures having a negative experience. After the BBC published, a spokesperson added that the suggestion the word “pregnancy” cannot be used on its platforms is incorrect, and that people can freely use it in posts.
Not only advertisers
Health creators and charities report posts removed, sensitivity filters applied, or reach suppressed — the practice known as shadowbanning. The campaign group CensHERship surveyed nearly 200 creators, clinicians, health brands, charities and educators; 96 per cent said they had met restrictions of some kind.
Athena Lamnisos, chief executive of the gynaecological cancer charity Eve Appeal, said some hashtags simply do not appear on Instagram — “one of the hashtags, unbelievably, was #vaginalcancer”. A labelled anatomical diagram posted to explain what changes to look for was pulled; an appeal was denied, then weeks later the post reappeared behind a sensitivity filter, the kind normally reserved for graphic imagery. Lamnisos said it was not an exaggeration to say life-saving health information is not reaching the people who need it most.
Deirdre O’Neill of the fertility testing company Hertility described a video in which a woman simply describes her endometriosis pain — no sales pitch, no company name. It was removed and classified as political content.
Why this keeps happening
Dr Carolina Are, a digital criminologist at the London School of Economics, says research shows AI is widely used to flag posts that might breach guidelines. The system misdetects, then stores that outcome and repeats it. Appeals are often automated too, which leaves people unable to put a question to anyone.
The result is self-censorship. CensHERship found the share of respondents censoring themselves rose from 53 per cent in 2025 to 65 per cent this year — writing “seggs”, using asterisks and emojis, or dropping subjects altogether. Co-founder Clio Wood called it damaging, because women’s health has historically been surrounded by shame, and this reinforces it.
Why this lands harder in Bangladesh
Bangladesh runs one of the world’s largest Facebook-native retail economies. Tens of thousands of F-commerce pages sell maternity wear, nursing products, sanitary goods, baby items and home-made food, and they are run overwhelmingly by women, many without a website, a shop or a payment gateway. The page is the business.
An automated restriction that an American brand survives with ten months of emails is, for a Dhaka seller, simply the end of their reach — there is no account manager to call, no English-language appeal that gets read, and no second channel. And because enforcement is tuned on English, a Bangla product description is judged by systems with far less training behind them, which cuts both ways: wrongly removed, or wrongly left up.
Two defensive habits are worth adopting now. Keep a customer list you own — phone numbers, a WhatsApp group, anything that is not a platform follower count — so reach suppression does not sever you from buyers. And keep a record of every removal and appeal, with dates and screenshots, because the only cases that ever get resolved are the ones somebody can document.
Related: fans training their algorithms to hide what they love.
