Why Great Content No Longer Works (MIT Research)

Search Engine Journal, citing MIT’s Work Analytics Lab “AI Labor Exposure Map” (65% of marketing-specialist tasks now automatable by current AI). A pointed take on the ai-search-shift that argues the standard response — generative-engine-optimization — may be the wrong game.

The argument

AI Overviews extract content to fuel AI answers, so users stop clicking to the source — “the great digital enclosure of publishing” (Rand Fishkin). Content becomes a commodity; “great content on your own site” no longer earns traffic because the value is captured at the answer layer.

The recommended response (note the tension)

• Ignore traffic. Build inimitable products. Shift from “great content on your site” to “great marketing on the platforms where your audience pays attention,” and toward the ~35% of work AI can’t replicate: original research, direct community access, irreplaceable human judgment.
• Collective action — withhold content / negotiate leverage (realistic mainly for large publishers).

Why it matters here

A second, contrarian answer to the ai-search-shift: where generative-engine-optimization and brand-depth-ai-recommendations say optimize to be recommended by AI, this says the channel is being enclosed — stop chasing it and differentiate on what AI can’t commoditize. That is a genuine strategic tension (see synthesis), not a contradiction of fact. (Caveat: SEO-trade op-ed citing one MIT dataset; the 65% figure is a labor-exposure estimate, not a measured outcome.)

Related

ai-search-shift · generative-engine-optimization · brand-depth-ai-recommendations

Tier note (re-graded 2026-08-04)

T4 → T2. The 2026-08-04 rule grades a page by what it rests on, and this one rests on a named third-party measurement — MIT’s Work Analytics Lab AI Labor Exposure Map and its finding that 65% of marketing-specialist tasks are automatable by current AI. That is independent analysis relayed, not the assertion-without-evidence that the rest of this spoke’s trade-press commentary runs on, and it is the only such page here.

The honest caveat: what this wiki holds is SEJ’s argument built on MIT’s number, not the map itself. Holding the primary source directly would be better, and is filed as a growth edge.