Fix “Promotional overlay on image” on Google Merchant Center

The image text overlay policy was tightened in 2024 and tightened again in 2026. The threshold isn't published officially, but operational evidence across accounts puts it at roughly 20% of the image area covered by text. The auto-reviewer over-flags during peak retail windows (Q4, BFCM) when manual review capacity is stretched, so creatives that pass in March may fail in November. The classifier looks at the entire image: badges in corners count, watermarks across the centre count, "100% Cotton" stamps over the product count.

What specifically triggers the flag: sale percentages ("SALE 50% OFF"), urgency claims ("LIMITED TIME"), product attributes baked into the image ("WATERPROOF", "100% COTTON"), brand watermarks larger than 10% of the image area, calls-to-action ("SHOP NOW"). What doesn't trigger: text that's a physical part of the product (book covers, garment text, product packaging photographed naturally), small corner brand logos under 5% of area, model release captions that don't promote.

The hand-edited-fix problem: removing overlay from 20 product images is feasible, removing it from 20,000 isn't. The scalable answer separates source-image policy compliance from creative branding. Keep image_link as clean source photography — no overlay, no badges, no watermarks beyond minimal corner branding. Apply promotional elements (sale badges, dynamic price tags, brand colour bars, conditional sale signals) at render time through a creative template that pulls data from the feed.

This is what the DPA Creative Editor handles. The template renders against every SKU automatically — change a price, the rendered creative updates next render pass. Add a sale badge to the template, every SKU's creative now shows the badge when sale_price is set. The source image stays policy-compliant across all channels; the rendered output carries the branding and promotion data. One template can render against 50,000 SKUs in under 15 minutes; manual badge work across the same volume takes weeks.

The auto-reviewer's seasonal aggression is worth planning around. October through December, the auto-reviewer is materially stricter — overlay that passes in October may flag in November. Q4 launches that depend on promotional creative should validate against current enforcement two weeks before the campaign goes live, not three days before. Most teams discover this when their BFCM creative gets flagged mid-campaign.

The appeal path for false positives is real but slow. Legitimate text that's a physical part of the product (book titles, garment text, product names physically printed on packaging) sometimes gets flagged. The appeal succeeds when you can document the text is physical-to-product, not added-in-post. Provide the original photography file and the manufacturer's product image showing the same text. Appeals typically resolve in 5-7 days; during the wait, the listings are suspended.

The strategic implication: this is the work where you separate "policy-compliant feed-grade imagery" from "branded creative-grade imagery". They're different assets with different requirements. Treating them as one breaks the moment policy tightens. The teams that have done this separation well are the ones who can ship creative refreshes across the full catalog in hours instead of months, because the rendering layer scales independently of the source-image layer.

The Audit feature catches overlay issues across Google, Meta, and TikTok (each has its own overlay policy with slightly different thresholds) — see the [DPA creative playbook](/blog/branded-dpa-creative-catalogue-ads-at-scale) for the full creative-layer strategy.