Meta Reviews the Ad Again After It Has Run
Meta's Indian enforcement against exploitative advertising is remediation after delivery, stated as policy in the company's own words. An apparel or beauty brand buying that inventory is purchasing a record of where its money went, not an assurance about where it will go.
Sir John Crabstone
Meta’s ad review runs twice. The first pass happens before an ad is served; the second happens at any point afterwards, which makes the second the binding one. An apparel or beauty brand buying Indian Instagram inventory is buying that second pass: an account of where its money went, produced after it went there.
The evidence for how the second pass performs arrived on 3 July. A BBC investigation ran a test account in India and logged roughly thirty paid advertisements promoting child sexual abuse material, routing users to Telegram channels that sold it for 99 rupees. When the BBC reported one of the adverts, Instagram replied within a day that “the advertiser’s ad does not go against our community standards”. The first pass had cleared it. So had the second.
Hours after the investigation ran, India’s technology ministry summoned Meta and gave it seven days to explain how the advertisements had been approved. That is the right question, and it is the one a brand cannot ask about its own placements. A ministry can compel an account of a decision; a media buyer receives the account the platform elects to publish.
Meta’s answer, posted four days later, describes the machinery without embarrassment. The system “automatically checks ads for policy violations before they run”. Ads then “remain subject to review and re-review at all times and may be rejected or restricted for violation of our policies at any time”. Read that second sentence as a media buyer rather than as a regulator. It promises that the platform may withdraw its approval after you have paid for the distribution it approved.
The same post cites 160,000 accounts removed in India in six months, carried across the Indian press as proof of enforcement. That figure belongs to a separate defence, detection of “suspicious off-platform links,” not to the ad-review pass just quoted. Meta has not said how many of those accounts were advertisers, or how many ads had run before removal.
An impression cannot be recalled; it can only be described.
Coverage has treated this as a child-safety failure and a regulatory exposure for Meta, and it is both. Treating the advertiser as a spectator is the error. Every brand running in India bought its placements from that review system, adjudicated to that standard, on that schedule.
This is where the brand safety apparatus sold to advertisers stops being reassuring. Meta’s inventory filter sorts the organic content an ad may sit beside into three tiers of sensitivity, and the most protective tier does its work by buying less. Third-party verification through Zefr covers Facebook Feed, not Instagram, so the tool most often cited as proof of oversight has no reach into the inventory this brief concerns. Suitability settings classify posts; they say nothing about the other advertisers clearing the same auction into the same feed, a set no brand running broad prospecting can inspect.
Six days after the blog post, Forbes argued that the failure was ongoing rather than resolved. Its evidence was the same advert the BBC had already documented, re-read as proof the pattern hadn’t stopped. Nobody has yet shown a fresh instance. Nobody has shown the pattern ended either.
The practical consequence lands in the brief. Enforcement of this shape produces a defensible file of dates and removals; it does not produce a guarantee. Adjacency has been sold to marketers as a setting, and on this inventory it is a probability with a report attached. A brand that wants certainty about what its advertising sits among in India is asking for a product Meta does not sell, and the seven-day notice from Delhi is the closest anyone has come to obtaining one.
Meta’s other July announcement makes the sequencing plain. On 16 July it began alerting parents when a teenager shows signs of distress in conversations with Meta AI, live in the United States, Britain, Australia and Canada, with the rest of the world promised “by the end of the year”. India sits in the rest of the world. Protection is sequenced by market; the inventory never was.
The alerts reach India by December. The impressions are on sale this morning.