Playbook · Paid Search

Paid Search Strategy for 2026

Search used to be a keyword game. In 2026 it is a signals game. Google's AI decides more of the query matching, the bidding and the ad copy every quarter, and your job has moved upstream to the inputs that steer it.

By Robert Rotuk · · 9 min read

Ten years ago a strong search account was a beautifully organized list of keywords, match types and bids. You could out-work a competitor by building tighter ad groups and adjusting bids more often.

That account still exists in plenty of advertisers’ heads, but not in Google’s product. In 2026 the system decides which queries match your ads, how much to bid in each auction, which headline to show and, increasingly, which landing page to send people to. Your leverage has not gone away. It has moved to the inputs the system learns from: your conversion data, your budgets and targets, your creative and your landing pages, and the guardrails you set.

This playbook covers how to structure Google Ads and Microsoft Ads for that reality, where automation earns its keep and where it needs a firm hand, and the new paid surfaces inside AI answers.

The short version

  • Feed the machine the right goal. Bid to the conversion that matters to the business (qualified lead, booked appointment, profit), not the one that is easiest to count.
  • Consolidate. Fewer campaigns with more data learn faster than many campaigns with little data.
  • Automate with guardrails. AI Max, Performance Max and broad match can find demand you would miss. Brand exclusions, negatives, URL rules and asset controls keep them honest.
  • Creative and landing pages are now search levers. The system writes and picks ads from what you give it.
  • Prove incrementality. Search is great at taking credit for demand that already existed. Test what it really adds.

What changed in 2026

The pace of change in Google Ads this year was relentless. The changes that matter most for strategy, from Google’s own announcements:

  • AI Max for Search kept expanding. Text guidelines (rules for what AI-generated text can and cannot say) opened to all advertisers in February. New planning and testing tools followed in August.
  • Legacy features are being folded into AI Max. Google said in April that, from September, features such as Dynamic Search Ads auto-upgrade to AI Max. If you still run DSA, plan for it.
  • Performance Max got more steering. Asset experiments arrived in April, and Google signposted more steering and reporting updates through the year.
  • Campaign total budgets for Search, Performance Max and Shopping (January) make fixed-period pushes, such as a sale or a launch, easier to run without daily-budget babysitting.
  • Ads in AI Mode gained new formats at Google Marketing Live in May, alongside demand-led budget pacing for Search and Smart Bidding Exploration expanding to Shopping.
  • Video creative inside Google Ads. Veo became available in March, lowering the cost of video assets for Performance Max and Demand Gen.
  • Faster ad approvals. Real-time policy reviews for responsive search ads cut review times from hours to seconds.

Outside Google, ChatGPT became a buyable channel. OpenAI added CPC buying to ChatGPT ads in May 2026, began rolling out a self-serve Ads Manager in beta, and launched a Conversions API and pixel for measurement.

Part 1: Get the goal right

Every automated bid strategy optimizes toward the conversion you tell it to value. Choose badly and the system will very efficiently deliver the wrong thing.

  • Use primary conversions sparingly. Only the actions that represent real value should be primary. Page views, scroll depth and “clicked phone number” belong in secondary goals.
  • Send back what happened later. For lead generation, import offline conversions from your CRM (qualified lead, booked appointment, closed deal) so Smart Bidding learns from quality, not form-fill volume. A clinic that bids on form fills gets cheap form fills. A clinic that bids on booked appointments gets patients.
  • Use values. Even rough values (“a booked consultation is worth five times a newsletter sign-up”) let value-based bidding prioritize properly. For e-commerce, bid on profit or margin where you can, not just revenue.
  • Protect the data. Enhanced conversions, Consent Mode and server-side tagging keep conversion signal flowing as browsers and privacy rules tighten. The Measurement & Incrementality Strategy covers this in depth.

Part 2: Structure for learning

The old instinct was to split everything: one ad group per keyword, one campaign per match type, one campaign per device. That starves automated bidding of data.

A sensible 2026 structure for most accounts:

Campaign Job Notes
Brand Search Protect your own name cheaply Exact and phrase match on brand terms. Exclude brand from everything else
Core non-brand Search Capture high-intent demand Themed ad groups, broad or phrase match with AI Max, strong negatives
Performance Max Find incremental demand across Google’s inventory Feed-based for e-commerce, lead-based with offline conversions for services
Demand Gen Create demand on YouTube, Discover and Gmail Judge it on assisted value and lift, not last click
Competitor (optional) Intercept comparison shoppers Separate budget, clear differentiation in the ad

Group by what changes the economics: different margins, different conversion values, different locations or different budgets. Do not split by anything the algorithm can already handle itself.

Part 3: AI Max and broad match, with guardrails

AI Max for Search combines broader query matching, AI-generated text and final URL expansion on top of a normal Search campaign. Used well, it finds valuable queries you never thought to bid on. Used carelessly, it spends money on loosely related traffic and sends people to the wrong page.

Treat AI Max as a test, not a switch. My approach:

  1. Turn it on in one campaign with clean conversion data, not across the account.
  2. Set brand exclusions so AI Max does not cannibalize your cheap brand campaign.
  3. Use URL rules and exclusions to stop final URL expansion sending traffic to blog posts, careers pages or anything that does not convert.
  4. Write text guidelines that rule out claims you cannot make. This matters in regulated categories such as healthcare and finance.
  5. Review search terms weekly for the first month, adding negatives and noting the new themes that perform.
  6. Compare against the baseline on cost per qualified conversion, not on volume.

The GAds Checklist covers the account fundamentals to get right before you switch it on.

Part 4: Performance Max without losing the plot

Performance Max decides channel, placement and asset combination across Search, Shopping, YouTube, Display, Discover, Gmail and Maps. Its strength is reach and learning. Its weakness is that it will happily take credit for brand searches and remarketing that would have converted anyway.

  • Exclude your brand using brand exclusions, unless you deliberately want PMax to cover it.
  • Use channel performance reporting to see where spend actually goes, and question any campaign where most conversions come from brand search or remarketing.
  • Feed it great assets. Several distinct images, real video (now cheaper with Veo inside Google Ads), and headlines that each say something different. Use asset experiments to test themes, not single words.
  • Use audience signals as hints, not targeting. Customer lists and website visitors help the system start in the right place. They are not hard limits.
  • For lead generation, offline conversions are not optional. Without them PMax optimizes toward whoever fills in forms most easily, which often means junk leads.

Part 5: Creative and landing pages are search levers now

Responsive search ads, AI-generated text and final URL expansion mean the system assembles your ad and chooses your page. So give it great material.

  • Headlines with distinct jobs: the offer, the proof, the differentiator, the objection, the call to action. Fifteen variations of the same phrase teach the system nothing.
  • Real proof in assets: reviews, ratings, prices, guarantees and credentials. Use sitelinks, callouts, structured snippets and images.
  • Landing pages that match the query and the ad. Message match still decides conversion rate. If AI Max can expand URLs, make sure every page it might pick is one you would choose. The CRO & Landing Page Strategy goes deeper.

Part 6: Microsoft Advertising

Microsoft Ads is easy to treat as an afterthought: import the Google campaigns and forget. That leaves money on the table.

  • Import, then adjust. Imports are a good start, but bids, budgets and audiences need their own tuning. Microsoft’s audience skews older and more desktop and business-heavy.
  • Use LinkedIn profile targeting, which is unique to Microsoft, for B2B: company, industry and job function on search audiences.
  • Run Performance Max on Microsoft too, with the same brand exclusion and asset discipline as on Google.
  • Watch Copilot and partner inventory. Microsoft places ads in its AI experiences and across its search partners. Check placement and quality reports regularly.

Part 7: Ads inside AI answers

Paid search no longer stops at the results page.

  • Google AI Overviews and AI Mode show ads drawn from your existing Search, Shopping and Performance Max campaigns. There is no separate campaign type to buy. Broad match, AI Max and PMax decide whether you appear, which is another reason to set their guardrails carefully.
  • ChatGPT ads are a separate buy. As of May 2026 you can buy on CPC through OpenAI’s partners or the self-serve Ads Manager beta. Start small, track conversions through the Conversions API or pixel, and treat early numbers as directional while the platform matures.

The strategic point: AI answers compress the research journey. Someone who asks an assistant to compare three providers may arrive ready to buy, or may never click at all. Organic AI visibility and paid placement have to be planned together. The AI Search Strategy covers the organic side.

Part 8: Measure what search really adds

Search captures demand brilliantly, which makes it brilliant at claiming credit. Three habits keep you honest:

  1. Separate brand from non-brand in every report. Blended ROAS hides the fact that brand search is cheap and mostly incremental to nothing.
  2. Run geo or holdout tests on major changes: a new PMax campaign, a big budget increase, turning on AI Max. Google’s own experiments and conversion lift tools help.
  3. Read assisted value, not just last click. Demand Gen and upper-funnel PMax work earlier in the journey. Judge them on lift and assisted conversions.

The weekly and monthly rhythm

Cadence What to do
Weekly Search terms and negatives, budget pacing, disapprovals, conversion tracking health, new query themes from AI Max
Fortnightly Asset performance, new creative into rotation, landing page checks
Monthly Channel and placement review for PMax, target adjustments, audience list refresh, brand vs non-brand trend
Quarterly Structure review, incrementality test, offline conversion audit with sales or front desk

A 90-day plan

Days 1–30: Foundations

  • Audit conversion actions: primary vs secondary, values, enhanced conversions and offline imports.
  • Split brand and non-brand reporting. Add brand exclusions to PMax and AI Max.
  • Consolidate thin campaigns and ad groups.
  • Plan for the Dynamic Search Ads to AI Max upgrade if you run DSA.

Days 31–60: Test automation properly

  • Launch AI Max in one campaign with URL rules, text guidelines and a clear baseline.
  • Refresh assets: distinct headlines, new images and at least one real video per PMax asset group.
  • Set up a geo or holdout test on your biggest automated campaign.

Days 61–90: Expand and diversify

  • Roll out what worked in the AI Max test. Retire what did not.
  • Tune Microsoft Ads separately: bids, audiences and LinkedIn targeting for B2B.
  • If your category has intent there, run a small, tracked ChatGPT ads test.

Common mistakes

  • Optimizing to the easiest conversion instead of the valuable one.
  • Letting PMax and AI Max eat your brand traffic and calling it growth.
  • Turning on every automated feature at once, so you cannot tell what worked.
  • Ignoring search terms because “the AI handles it”. It handles matching, not judgment.
  • Starving campaigns of data by splitting budgets across too many campaigns.

The bottom line

The paid search manager’s job in 2026 is less about pulling levers and more about deciding what the machine optimizes for and what it is not allowed to do. Get the conversion signal right, consolidate, feed strong creative and pages, set firm guardrails, and test what the platform really adds. Do that, and automation stops being something that happens to your account and becomes the most productive person on your team.

Last reviewed November 16, 2026. Google ships changes almost weekly. I update this playbook when the strategy, not just the interface, changes.

Sources and further reading

  1. Google Ads product announcements — Google, accessed November 2026
  2. About Performance Max campaigns — Google Ads Help, accessed November 2026
  3. About Demand Gen campaigns — Google Ads Help, accessed November 2026
  4. New ways to buy ChatGPT ads — OpenAI, 2026-05-05
  5. Google Ads updates 2026, by theme — Hawky, 2026-07-23