Automated comment management was followed by more registrations per click in one advertising campaign and more purchases in a second test, while another experiment found that revealing certain hidden comments could weaken trust in a brand.
Jiwoon Park and Julian De Freitas examined moderation and engagement around social media advertising in a 2026 Harvard Business School working paper. The research is preliminary and should not be treated as a guaranteed return on a moderation tool.
One of the field tests involved an automated service supplied by a comment-management business. Results from the specific advertisers, platforms and campaigns cannot establish what every company would earn from a similar system.
Two different field comparisons
In a financial-services campaign, completed registrations per ad click increased from 1.67% to 1.94% after automated comment management was introduced, according to the Harvard Business School AI Institute. The campaign ran for five months before and five months after the intervention.
The before-and-after design leaves room for other changes over time. A second comparison ran campaigns concurrently on Facebook and Instagram, with automated management enabled for one condition and disabled for the other.
Each condition spent roughly $10,000. The managed campaigns produced 22 purchases and the comparison campaigns 15. Return on advertising spend, a measure of sales revenue divided by ad spend, was 0.678 and 0.459 respectively. Both figures are below one, so the reported sales revenue was below advertising expenditure before other costs.
The low number of purchases makes the difference sensitive to individual transactions. The field system combined hiding harmful comments with replies, so the campaign results cannot isolate the effect of each action.
Moderation and customer trust
In separate online experiments, participants evaluated a fictional brand in controlled ad settings. Moderating harmful comments was associated with higher stated purchase intentions, while the tested engagement condition did not produce a comparable statistically significant effect.
Stated intention is not a completed sale. Another experiment found less favorable responses when participants were told that brand-directed hostile comments had been hidden. The pattern differed for comments involving material such as hate speech or profanity and when a brand disclosed its moderation policy.
Removing spam or threats serves a different purpose from hiding legitimate criticism. Businesses can publish clear rules, review automated mistakes and give customers a way to raise unresolved complaints.
What advertisers can measure
A company considering moderation can compare similar campaigns over the same period, track the kinds of comments removed and measure purchases, customer-service issues and costs. It should also check platform rules and applicable consumer-protection requirements.
Our earlier coverage of prospecting and retargeting ads examined another reason to test incremental results: a platform’s attribution report does not necessarily show which purchases a campaign actually caused.
