AI Fatigue 2026: DuckDuckGo No-AI Search

On June 1, 2026 DuckDuckGo expanded access to its no-AI search option after traffic to the feature surged through the spring. The product change is small. The signal it carries is large. A privately held search engine has concluded that the fastest-growing slice of its user base is people who specifically want fewer AI features, not more, and is reworking its surface to make that preference one click away.

The DuckDuckGo move sits inside a broader cultural and commercial pattern that has been hardening through 2025 and into 2026 — AI fatigue. This is the product-leadership read.

The DuckDuckGo signal#

DuckDuckGo has run a no-AI search mode since 2024 — a setting that returns clean web results without AI-generated summaries, AI-rewritten snippets, or AI-recommended content. The June 2026 change makes that mode easier to reach from the default surface. The product team’s framing is that user research showed enough demand for a clean web-search experience to justify foregrounding it in the UI.

The interesting question is not why DuckDuckGo built the feature. It is why a meaningful fraction of search users now actively prefer a result page without AI in it, and what that preference says about the consumer market AI products are being sold into.

The numbers behind the surge#

Public traffic data on DuckDuckGo is limited, but adjacent indicators line up. Mozilla Firefox added a no-AI default option in 2025 and reported uptake in the user-research panels that informed the change. Brave’s anti-AI-summary toggle saw similar interest. Reddit communities organised around AI-free workflows have grown. The pattern is not a small fringe — it is a coherent consumer segment that product teams now have to account for.

The AI slop problem#

A large fraction of the consumer AI experience over the last two years has been low-quality. Auto-generated SEO articles, AI-rewritten product reviews, AI-summarised news that gets the facts wrong, AI customer-service flows that loop indefinitely, AI-generated cover art that all looks the same, AI-padded LinkedIn posts. The collective term that emerged through 2025 is AI slop, and the cultural rejection of it has moved from niche complaint to mainstream consumer pattern.

The slop is not the model’s fault in a narrow technical sense. It is the consequence of every consumer-facing product team being asked to ship AI features fast, with weak quality bars and weak feedback loops. The user-experience cost has been pushed onto the user, and the user is now telling the market that the cost was too high. The market is starting to listen.

Google AI Overviews and the publisher revenue collapse#

The clearest commercial signal of AI fatigue is what has happened to Google AI Overviews. Google launched AI Overviews as a default feature on search results in mid-2024. By late 2025 the feature was on a meaningful share of queries. Publisher CTR data from across the industry showed declines on queries where an Overview appeared — declines that varied by sector but ran into double-digit percentage drops on informational queries.

The publisher economics consequence has been severe. Independent publishers that depended on search referral traffic for ad revenue have seen quarter-on-quarter revenue declines tied directly to Overview presence on their topic terms. The structural argument is now widely accepted: when a search engine answers the user’s question on the result page, the user does not click through, and the website that produced the underlying answer does not get the visit.

Google has adjusted Overview presentation through 2025 and 2026 in response to publisher pressure and to a flattening of user engagement on Overview-heavy result pages. The DuckDuckGo no-AI surge sits in this context — a part of the consumer base has decided that the Overview model is worse for them, not better, and is voting with their default search engine. For the deeper search-disruption picture see the Google search AI disruption piece and AI search versus traditional search.

Apple’s slower posture#

Apple’s response to the generative-AI wave has been notably more cautious than the rest of the industry. The Apple Intelligence feature set was announced in 2024 and rolled out through 2024 and 2025 with explicit opt-in surfaces, on-device-first inference for routine queries, Private Cloud Compute architecture for the larger workloads, and a Siri overhaul that was delayed multiple times into 2025 and 2026.

Apple’s framing of the delay was technical. The product-leadership read is broader: Apple chose not to ship AI features that would degrade the user experience for the sake of the launch calendar, and the market response to that posture has been positive. Customer-satisfaction data on Apple Intelligence — where it shipped — has held up better than the comparable data on Galaxy AI on Android and on the Microsoft Copilot integrations across Windows. The slower posture has aged well. For a fuller treatment see the Apple Intelligence strategy piece.

The Siri delay specifically is the lesson. A consumer voice assistant that gets the answer wrong, or that gets the action wrong on a user’s calendar, is worse than a voice assistant that does not exist. Apple appears to have understood that before its competitors did.

The anti-AI product positioning trend#

A small but coherent set of consumer products is now positioning around the absence of AI as a feature. No-AI laptops marketed at users who want a clean operating system without Copilot integration. No-AI camera apps that emphasise unedited image fidelity. No-AI writing tools that explicitly do not run inference on user text. Newsletter platforms that explicitly do not train on subscriber emails. Indie podcasts that label themselves AI-free in titles and show notes.

This positioning is a market-research signal. The product teams shipping these features have looked at consumer-research data and concluded that there is a paying audience for AI absence. The audience is not the whole market. It is large enough to be a commercial opportunity that did not exist three years ago.

Why this matters for B2B product#

The B2B implication is the most important one for our enterprise readers. If a meaningful fraction of consumer users now prefer fewer AI features, and product teams selling to those users are responding, the same dynamic is starting in B2B. Knowledge workers who are tired of Copilot summarising their meetings, of AI rewriting their emails into stilted prose, of AI assistants interrupting their work to suggest improvements, are starting to ask their IT teams to turn the features off. IT teams who have rolled out AI features under enterprise licences are starting to see usage data that does not justify the per-seat cost.

UX patterns that respect user choice#

The product pattern that the AI-fatigue moment is pushing toward is opt-in AI rather than always-on AI. Specifically:

AI features off by default for users who have not actively enabled them, rather than on by default with an off switch hidden three menus deep.
Surface-level visibility of when AI is active on a given workflow, so the user is not surprised by AI-generated content masquerading as their own.
One-click escape hatches from AI-driven flows back to the underlying tool, especially for search, writing, and customer-service contexts.
Honest labelling of AI-generated output, both for the user’s awareness and for the broader trust position of the product.
Workflow-level rather than feature-level toggles, so a user can keep AI in some contexts and refuse it in others, without an all-or-nothing decision.

These are not radical patterns. They are the patterns that good consumer-product teams already used for any feature that some users wanted and others did not — push notifications, location services, analytics. AI is being asked to live by the same rules.

The regulatory implication#

The consumer-protection regulators are watching the AI-fatigue pattern carefully. The argument is straightforward: if a meaningful fraction of consumers do not want AI on their results, and the platform makes it difficult to turn off, that is a deceptive-pattern claim under the consumer-protection statutes most jurisdictions already enforce. The EU’s Digital Services Act has provisions about default settings and user choice that map directly onto this argument. The FTC has been signalling through 2025 that it is preparing similar enforcement under its Section 5 authority.

Product teams who design AI features as always-on with hidden opt-outs are running into the same regulatory wall that cookie banners ran into a decade ago. The path through is to ship opt-in defaults from the start.

What product leaders should take from this#

Three reads.

First, AI features are no longer free. Two years ago every product team got to ship AI for the marketing benefit. That window has closed. AI features now have to clear a quality bar and a user-consent bar, and the cost of failing either bar is loss of trust that compounds across the product surface.

Second, the absence of AI is now a product feature in its own right. There are paying customers who specifically want clean, predictable, non-AI workflows, and the addressable market for that posture is large enough to be commercially viable.

Third, the opt-in default is the right default. The opposite framing — always-on with an opt-out — is becoming a regulatory liability and a trust liability simultaneously. Get ahead of both by shipping AI features with an opt-in surface and a visible label.

Where pdpspectra fits#

We work with product and engineering teams on AI integration architecture, including the UX-and-governance side of when to ship a feature, when to keep it opt-in, and how to instrument adoption so the data tells you whether users actually want it. Our AI and LLM integration practice covers the consumer-product and B2B-SaaS side of this work and we routinely review feature-rollout plans with client product teams.

The AI-fatigue moment is a signal worth taking seriously when planning the next feature rollout. Talk to us if you want a second opinion on the AI features your roadmap is about to ship.