Contextual advertising — placing ads based on what a page is about rather than who is reading it — was the industry standard until behavioural targeting displaced it. For most of the last decade it has been treated as the option you fall back on when addressability fails.
That framing no longer matches how the stack works.
Why the position changed
Two developments moved contextual back to the centre.
Google discontinued the Privacy Sandbox initiative in October 2025, retiring core APIs including Topics, Protected Audience and Attribution Reporting after low adoption and continued regulatory pressure. Separately, third-party cookies were never deprecated — but their reliability declined, with match rates on authenticated traffic settling near 47% against roughly 68% in the cookie era.
The result is that programmatic buyers now run a hybrid stack: authenticated identifiers such as UID2, ID5 and RampID, contextual signals, and clean-room measurement. Contextual is listed as a component, not a contingency.
What is different from the old version
Contextual targeting in 2026 is not keyword matching against page text. Classification has improved substantially — page-level semantic analysis, sentiment, and brand-safety scoring operate at a level that keyword lists never reached.
This matters because the historic objection to contextual was imprecision: an article mentioning a car crash would serve car ads. Modern classification distinguishes subject from incidental mention.
Where contextual outperforms
Upper-funnel campaigns. Identity resolution adds little when the objective is reach against an interested audience. A reader of a cycling article is a valid cycling audience regardless of whether they resolve to an ID.
Environments where addressability is weakest. The 47% match rate is an average. Some inventory performs far worse, and contextual signal quality does not degrade there.
Categories with regulatory sensitivity. Health, finance and anything touching children carry restrictions on behavioural targeting that contextual placement sidesteps entirely.
Where it does not
Contextual cannot do frequency capping across sites, sequential messaging, or suppression of existing customers. Those need identity, and no amount of page classification substitutes.
The honest position is that contextual and identity-based targeting solve different problems. Treating either as a replacement for the other produces disappointment.
What to change
Split budgets by objective rather than by method. Upper funnel toward contextual, lower funnel toward addressable inventory.
Re-test contextual if your last evaluation predates 2023. The classification technology has changed enough that old benchmarks are not informative.
Stop describing it as a fallback internally. The language shapes budget decisions, and the data no longer supports it.
Programmatic spend reached an estimated $755 billion in 2025 and is projected to clear $821 billion in 2026 — roughly 90% of global digital display. The money did not move because addressability wobbled. It redistributed across methods.
