The keyword used to be the unit of PPC. You picked the phrase, set the match type, wrote the ad, and the searcher typed something close enough to trigger it. AI search has broken that loop: people now ask long, conversational questions instead of typing two-word phrases, and platforms match ads to meaning rather than to exact words.
This changes what a keyword is for: it is becoming a signal you feed the machine — a starting point for matching — rather than a precise trigger you control. This guide covers match types, bidding, search-term mining, negatives, landing pages and measurement, and ChatGPT ads, where keywords do not exist at all.
Key Takeaways
- AI search queries are longer — Google says AI Mode queries run about three times longer than traditional searches — so intent, not exact phrasing, is the targeting unit.
- In AI Overviews and AI Mode, Google matches ads directly to the query or contextually to the query plus the AI answer. Broad match, AI Max, Shopping and Performance Max qualify for both; exact and phrase match only for direct intent.
- Broad match only works with conversion-based Smart Bidding and trustworthy conversion data. Keywords are a seed for the machine, not a fence around it.
- AI Max is an optimisation layer inside existing Search campaigns, not a new campaign type. It needs conversion-based Smart Bidding, and fails on budget-limited campaigns or manual bids.
- ChatGPT ads have no keywords — only context hints, plain-language descriptions of the conversations where an ad fits. OpenAI states hints are not exact-match keywords and guarantee no delivery.
- You cannot target or opt out of AI Overviews or AI Mode separately; results are folded into normal campaign reporting. Measurement means reading the search terms report, the AI Max source column, and your own conversion data.
What Actually Changed About the Keyword?
The keyword did not disappear, but its job changed. It used to be a trigger: type the phrase, see the ad. Today it is a signal the system uses to understand what you sell, who you sell it to, and what a conversion is worth.
The change starts with how people search. Google says AI Mode queries run about three times longer than traditional searches. Instead of “plumber near me”, someone asks which plumber can fix a burst pipe tonight, what the call-out fee is, and how long the job takes. That question carries the same commercial intent, but it will never match a two-word keyword exactly. Meanwhile Google’s AI reads the meaning of a query together with your ads, landing pages and keywords to decide whether your ad fits. The match is semantic, not literal: a keyword list is still a starting point, but it no longer defines the boundary.
The old craft was building exhaustive keyword lists; the new craft is feeding the system clean signals — less time predicting every phrase a buyer might type, more time making sure the machine understands your business.
Why Can’t Exact Match Reach the New Queries?
Exact and phrase match still work, but only within a narrower slice of search than they used to cover. Google’s own help pages draw the line clearly, and it matters for any advertiser still running a mostly exact-match account.
In AI Overviews and AI Mode, Google matches ads two ways. Direct matching: the search itself expresses clear commercial intent, such as a search for a product type by name. Contextual matching: someone asks a longer, exploratory question, and commercial intent emerges from the AI answer’s suggested next steps. Exact and phrase match are eligible for direct matching, but Google says they are not eligible for contextual matching on broader exploratory questions.
That is the gap: exploratory questions are where AI search is growing and where buyers increasingly start. Someone researching why their pool turned green is not searching for a pool vacuum — but the AI answer may suggest one, and a contextually matched ad can be there. An exact-match-only campaign cannot reach that moment. Keep exact match for your highest-intent searches — but an account built only on exact and phrase match is blind to half of search.
How Do Broad Match and Smart Bidding Work Together Now?
Broad match and Smart Bidding are a pair, and in the AI era they are effectively one system. Google’s guidance is consistent: broad match should run with conversion-based Smart Bidding, because the bidding is what turns a wide net into a profitable one.
Broad match today means your keyword can match searches related to its meaning, including searches that do not contain its words at all. That sounds reckless until you understand what Smart Bidding adds: at auction time, the system weighs the query, context, device, time, and above all your conversion history, and bids toward the searches likely to convert. The keyword opens the door; the bidding decides who walks through it.
That is why conversion data is the real foundation: if your tracking counts the wrong things — newsletter signups treated the same as booked jobs — Smart Bidding will optimise toward the wrong outcomes and broad match will amplify the error. Before widening match types, audit what counts as a conversion, then give the test a defined window, a success measure tied to real conversions, and budget to learn.
What Is AI Max, and When Should You Switch It On?
AI Max is Google’s bundle of AI features for Search campaigns — and the first thing to understand is what it is not. It is not a new campaign type, but a continuous optimisation layer you switch on inside existing Search campaigns; new Search campaigns have it on by default.
It has two main features. Search term matching expands your reach beyond your keyword list, using broad match and keywordless technology to find relevant queries you never listed. Asset optimisation tailors creative at scale: text customisation writes ad copy from your ads and landing pages, and final URL expansion sends traffic to a more relevant page. Google’s help pages attach clear conditions: the features need conversion-based Smart Bidding — Maximise Conversions, Maximise Conversion Value, Target CPA or Target ROAS — and search term matching will not work with manual bidding at all. AI Max also fails on budget-limited campaigns; Google shows an alert.
Test it first on one or two campaigns where tracking is solid and the budget is not constrained, using Google’s AI Max experiment format. Watch the search terms report, which shows “AI Max” as a match type with a source column for broad-match versus keywordless matches. If the incremental queries are relevant and conversions hold, roll it out; if they are noise, fix your signals first.
How Do You Mine Search Terms and Handle Negatives When the Machine Does the Matching?
Search-term mining used to be a harvest: find the exact phrases people typed, add the good ones as keywords, negative the bad ones. In an AI-matched account the system deliberately reaches beyond your list — and negatives change with it, from precision tools into one of the few hard boundaries you can still set.
Start in the search terms report. For AI Max traffic it shows “AI Max” as the match type, and a source column distinguishes broad match expansion from keywordless matching. Broad-match expansions usually stay close to your keywords, while keywordless matches reveal what the system learned from your landing pages — if that column is full of irrelevant queries, your pages are telling the wrong story. Mine for three things: new buying language, patterns over single phrases, and the queries that convert.
The discipline that changes is restraint: add keywords only for a distinct intent deserving its own ad copy or page. Build negatives in layers — structural ones first, then add from the search terms report weekly during a test, monthly once it settles. But negatives are a scalpel, not a wall: hundreds of them means the signal is wrong. Fix the signal first.
What Changes When AI Picks Your Landing Page?
Final URL expansion is the AI Max feature advertisers underestimate most: it takes a decision you used to make — which page an ad sends people to — and gives it to the system. Google says it will only send traffic to URLs that are query-relevant and themed to your ad group.
Every page on your site is now a potential landing page — stale offers, thin category pages, outdated prices and discontinued services can all receive paid traffic. Before turning it on, audit the pages it could reach: use URL exclusions for pages that should never receive ads — expired promotions, thank-you pages, login pages — and URL inclusions to steer it toward specific pages. Landing-page relevance also becomes a keyword strategy issue: the page must answer the query directly, and a bad page teaches Smart Bidding the wrong lesson about which queries are worth bidding on.
One technical catch: AI Max can generate dynamic landing pages, and tracking templates must be configured compatibly — Google warns a static tracking URL without the {lpurl} tag can cause 404 errors. Test that first.
How Do You Advertise Where There Are No Keywords at All?
ChatGPT ads are the clearest proof that keyword strategy is becoming intent strategy. There is no keyword targeting in ChatGPT ads. Instead, advertisers write context hints: plain-language descriptions, set at the ad group level, of the conversations, topics or keywords where their product or service may be relevant.
OpenAI is blunt about what hints are: they help guide ad matching but are not exact-match keywords and do not guarantee delivery in specific conversations. They are not targeting rules and cannot enforce geographic limits, schedules or exclusions — those live in campaign settings. The system reads the whole conversation, weighing its context and intent, the ad’s landing page, title and copy, the hints, and the targeting selections. Writing hints is a different skill from writing a keyword list: a keyword list rewards short, exact phrases, while a hint rewards a clear natural sentence about one buying situation — what you offer, who it helps, and when it is useful. OpenAI’s own example is plain words — cushioned everyday running shoes for beginners training for their first 5K — not a term list.
Campaigns buy on CPM or CPC — conversion campaigns on Target CPA, click campaigns on Target CPC. OpenAI moved new campaigns to target-based bidding from November 9, 2026, while existing campaigns kept earlier bidding until the start of 2027. Ads appear below ChatGPT responses, labelled as sponsored, only to Free and Go users. Our ChatGPT ads guide covers the format and measurement stack in more detail.
How Do You Measure Whether AI Is Surfacing Your Ads?
Measurement is where the AI shift is most uncomfortable, because the reporting does not keep up with the matching. Google is clear: you cannot target AI Overviews or AI Mode separately, you cannot opt out of them separately, and their results are included in aggregate in your normal campaign reports. There is no separate “AI Overview” segment.
So measure indirectly, and honestly about what you can and cannot see. The search terms report is your main window: which queries your ads matched, the AI Max match type and source column, and whether the share of long, conversational queries is growing — that shows whether AI matching reaches the new query shapes. For the business question — is this working? — the answer must come from your own conversion data. Compare conversion quality, not just volume: are leads from AI-matched queries as likely to become customers as exact-match leads? Segment your CRM or analytics by landing page or campaign, and use experiments rather than before-and-after comparisons — performance moves for a dozen reasons at once.
The same applies to ChatGPT ads: judge it on its own reporting trend plus the referral traffic you can see, since chatgpt.com referrals undercount the channel. One habit from the organic side transfers: our guide to tracking your citations in AI Overviews explains why AI visibility has to be sampled manually.
FAQ
Do I still need keywords at all?
Yes — as a seed signal, not a trigger. Keywords still tell Google what you sell and where to start matching. Keep a focused list covering your products and buyers’ intent, and let broad match and AI Max extend from there.
Should I switch all my campaigns to broad match?
No. Keep exact and phrase match for your highest-intent searches, where control matters most. Use broad match and AI Max as a discovery layer in separate campaigns with their own budgets, so you can see what AI matching adds without risking the proven core.
Is AI Max worth turning on for a small account?
Conversion volume, not account size. AI Max needs conversion-based Smart Bidding, which needs reliable conversion data to learn from. A small account with clean tracking and steady conversions can test it; one with thin or noisy data should fix measurement first — AI Max amplifies whatever the data says.
Can I see how my ads perform inside AI Overviews specifically?
Not separately. Google folds AI Overview and AI Mode results into normal campaign reporting in aggregate, and you cannot target or opt out of those placements. Judge it with the search terms report, the AI Max source column and your own conversion data — no placement-level report exists.
How are ChatGPT ads different from Google Ads for keyword strategy?
Google Ads still starts from keywords, even with AI matching. ChatGPT ads have no keywords at all — you write context hints describing the conversations where your ad fits, and OpenAI decides relevance from the conversation. Run them separately: keyword-led campaigns on Google, situation-led ad groups on ChatGPT.
Bottom Line
AI did not kill the keyword. It demoted it — from the thing that decides whether your ad shows, to one signal among several that the machine weighs. The advertisers who adapt are not the ones with the longest keyword lists, but the ones with the cleanest signals: keywords that describe real buying intent, conversion data that reflects real results, and landing pages that answer real questions.
Start with one campaign: pair broad match with conversion-based Smart Bidding, test AI Max with an experiment, mine the search terms report for themes, and build one ChatGPT ad group around a single situation. Measure on conversions that matter, not clicks that flatter, and make it a weekly habit — search terms on Monday, landing pages midweek, conversion quality on Friday. The keyword is no longer the strategy — it is the raw material.

