Google Ads + AI glossary

Ad Strength grades whether your responsive search ad has enough varied assets. What moves the label, when to ignore it, and how AI checks it account-wide.

Ads Advisor is Google's built-in AI agent for Google Ads: what it does, the incentive question it cannot escape, and how outside agents differ.

Google now places ads inside AI-generated answers. Which campaigns are eligible, how selection works, what reporting exists — confirmed versus still unknown.

Agentic AI plans and executes multi-step work with tools. What that means in advertising, the autonomy spectrum and the claims that deserve skepticism.

AI hallucination in PPC has a shape: convincing metrics with no tool call behind them. How grounding changes the equation and verification closes it.

AI Max is an opt-in layer on Search campaigns: expanded query matching, generated ad text, landing-page swaps. What each feature does and what control remains.

What an asset group is in Performance Max, how it differs from an ad group, how to structure one by audience story, and how AI helps build and review them.

What the auction insights report shows, its six metrics, what it cannot tell you, and how a monthly AI pull turns it into radar.

Audience signals seed Performance Max's learning — they do not restrict reach. What signals do, what makes a strong one, and how AI checks who PMax reached.

An audit trail records every call an AI made — tool, account, success or refusal, reason, timestamp. Why AI access to ad spend is unmanageable without one.

Google paces spend to daily budget times 30.4; your finance team paces to the calendar month. Budget pacing reconciles the two clocks before the invoice does.

Google Ads change history logs what changed, when, by whom and through which tool for two years — and it records AI-made changes exactly like human ones.

Conversion actions define what Google Ads counts as success. Primary vs secondary, counting and attribution settings, and how AI audits them in one pass.

Enhanced conversions send hashed first-party data with your tags so conversions lost to cookie limits get matched to signed-in users. How the two flavours work.

What GAQL is, how a query is built, where it differs from SQL, and how to sanity-check queries an AI writes against your account.

Human-in-the-loop means named actions block on named humans — the rest is theatre. Where checkpoints belong in ads work and how to avoid approval fatigue.

What impression share measures, how lost-to-budget differs from lost-to-rank, and when chasing a bigger share is the wrong move.

Exact, phrase and broad match as they behave in 2026 — same-meaning matching, the control spectrum, and how to audit what actually matched.

What a Google Ads manager account (MCC) is, how the hierarchy and access levels work, and how to scope what an AI assistant can reach.

What an MCP client is, which AI apps support MCP connectors today, and why approval behaviour is the client feature that matters most for ads.

What an MCP server is, what tools, resources and prompts mean, and how to judge any server — including a Google Ads one — by reading its tool list.

What the Model Context Protocol is, the integration problem it solved, and why it lets one Google Ads connector work in five different AI clients.

What n-gram analysis is in PPC, why fragments beat whole queries for finding waste, and how AI turned a script ritual into a conversation.

What the Google Ads OAuth consent screen actually grants, what the adwords scope can never reach, and how to revoke access in seconds.

Your search terms report is text strangers wrote — and your AI reads it. The ads-specific injection vector, what it can and cannot do, and the mitigations.

What Quality Score measures, how its three components relate to Ad Rank and CPC, and why the components beat the 1–10 number as a diagnostic.

What a refresh token is, how it differs from an access token, where connector vendors store it, and the custody questions worth asking.

Remote vs local MCP servers: what each setup demands, who holds credentials in each model, and when running your own server is still right.

What a responsive search ad is, how Google assembles it per auction, and how to write headlines as parts for a machine to recombine.

What the search terms report shows, what Google's privacy thresholds hide, and the waste–harvest–drift reading that finds money in it.

A seasonality adjustment tells Smart Bidding to expect a short, sharp conversion-rate change. When to use one, when it backfires, and how AI sizes it from data.

A shared budget lets Google move money to whichever campaign performs best today. When that pooling helps, when it inverts priorities, and how AI reviews it.

What Smart Bidding is, which strategies count, what auction-time bidding sees that you cannot, and what remains your job around it.

What Target CPA is, how to derive an honest target from close rates and deal value, and how to ramp it without resetting learning.

What Target ROAS is, the formula behind it, how to set a target from margin maths, and why raising the target lowers volume by design.

The model never touches your account — it emits a structured request that software validates and executes. Why that separation makes ads AI auditable.