Generative UI hits AI Overviews; Microsoft rewires Search bidding

Audit your tool and calculator pages against the queries they rank for — anything Google can trivially generate in-SERP is now a candidate for zero-click cannibalization. Prioritize pages where the value is proprietary data, gated workflows, or transactional intent that a generated widget can't complete.

If your AI-search strategy has quietly relied on ranking Reddit threads as a proxy for ChatGPT visibility, pull your citation logs from the last 30 days and see what replaced Reddit for your queries — that's your new target surface.

Your organic-traffic-is-down deck is incomplete without an AI Mode segment. Add the extraction to your monthly reporting before a client asks why total clicks are falling while impressions climb.

Don't diagnose ranking movement this week without checking whether the spam update or the AIO layout change is the actual cause — they're overlapping, and confusing the two leads to the wrong remediation.

For publisher and content-driven clients, add the Preferred Sources button to high-intent pages this sprint — it's a first-party signal opportunity that isn't dependent on ranking, and it costs nothing to test.

If Microsoft is a meaningful channel, stand up target CPA/ROAS strategies on parallel campaigns now so you have 30+ days of conversion data before Oct 1 — cold-starting automated bidding in Q4 with a broken conversion setup is how holiday budgets die.

Run the A/B test before your next QBR — 'AI Max lifted conversions' is not a defensible claim without a controlled experiment, and now you have Google's own framework to run one.

Test AI Max on a single campaign with brand exclusions and URL expansion controls locked down before turning it loose across the account — the URL expansion feature is where irrelevant landing pages show up in reporting.

Before you approve any 'agent-optimized' line item, get the vendor to define — in writing — what counts as an impression and how a conversion is attributed. If they can't, it's a research budget, not a media buy.

For clients on SGTM who are staring at rising container-instance bills, this is the tuning knob. Route pageviews client-side, keep purchases and lead events server-side — most of the cost is in the volume events, not the ones that matter for attribution.

If you run acquisition on TikTok for anything remotely youth-adjacent, revisit your audience config and creative — expect tighter platform-side restrictions on targeting and data collection to follow the settlement.

Ask your ChatGPT rep — or your agency — what the buying interface, targeting, and measurement look like before committing budget. The measurement gap on AI-agent ad exposure is real and unsolved.