In-Market Audiences in Google Ads: What You Are Actually Buying

What Google infers, how fresh it is, and the one setting that decides whether you narrow reach.

Most accounts switch in-market segments on because the interface suggests them, then never work out what they bought. This page is the audience layer of our Google Ads guide.

An in-market segment is a guess — Google's guess, from Google's data, about whether someone is in a buying window for a category Google defined. You cannot see who is in one, you cannot audit the guess, and you cannot add a category that does not exist.

That makes them a reporting instrument first and a targeting control second — the opposite of how most accounts use them. What follows: what the inference implies, how in-market differs in kind from affinity and custom segments, where the segments work, and the one setting that decides whether you have narrowed your reach or just asked for a report.

Every Google claim below was fetched from Google Ads Help on 28 September 2026. Where a page shows no last-updated date, we say so.

What are in-market audiences?

In-market audiences are Google Ads audience segments grouping people Google judges to be actively considering a purchase in a specific category. Google's "About audience segments" Help page defines them as reaching "users based on their recent purchase intent" — people "researching products and are actively considering buying a service or product like yours".

Membership is inferred, not declared. The same page says audiences are built from "people's activity using Google products and third-party websites, or estimated based on content certain groups of people are likely to be interested in".

Two layers of inference, then: what Google observed, and what Google concluded from it. You are buying the conclusion.

That page shows no page-level last-updated date — only "Apr. 28, 2023" against its health-related sections. Read 28 September 2026.

Audiences or segments? Google's current wording

Google's current term is audience segments. Its documentation states that the old "audience types" — custom, in-market, affinity — are now audience segments, and that "remarketing" is now "your data". Google's own pages still use "audiences" and "segments" interchangeably, so both words are current.

Three renames that break older tutorials:

• "Remarketing lists" → your data segments, the naming our remarketing explainer uses.
• "Custom intent" and "custom affinity" → one control, custom segments, taking keywords, URLs and apps.
• "Similar audiences" → gone; seed-list modelling now lives as Lookalike segments in Demand Gen, per our lookalike audiences post.

In-market vs affinity vs custom segments

These three are not three flavours of the same thing. They differ in who does the inferring and how fresh the signal is — and that is the whole basis for choosing between them.

One availability trap: in-market segments run on Search, Display, Video, Demand Gen and Shopping, but custom segments do not run on Search or Shopping — Google lists them for Display, Gmail, Demand Gen and Video only. So on a Search campaign, in-market is the inferred-intent option you actually have. Google's custom segments page shows no last-updated date; both pages read 28 September 2026.

What the inference actually implies

Three consequences follow from "inferred from recent behaviour" — and they are why these segments disappoint accounts that treat them as a customer list.

1. The signal is recent, and it goes away. Someone in a buying window this week may not be in one next month. So in-market is a poor base for a long nurture sequence, a reasonable one for capturing demand now — and a segment that performed in March is no promise about October.
2. The categories are Google's taxonomy, not your market map. You pick from a fixed list. If your product sits between two categories, or under one built for a different market, no amount of bid work fixes the mismatch — that is the case for a custom segment instead.
3. You cannot see who is in one, or why. No membership list, no per-person reason, no audit trail — only aggregate performance in the audience report. Google publishes no accuracy figure for the inference, so the only evidence you have is how the segment behaves in your own account.

That last point is why the observation setting matters more than the segment choice.

Where in-market segments can be used

In-market segments run on Search, Display, Video and YouTube, Demand Gen and Shopping — but not with the same job in each place, and on Performance Max they are not a target at all.

• Search. The query does the targeting; in-market rides alongside as a layer you observe, which is what Google recommends here. It does not replace query control — that is your match types and negatives.
• Display. Targeting genuinely narrows: Google states that "generally each time you add a new layer of targeting, you narrow the potential reach of your ads".
• Video and YouTube. Google's Video targeting page describes them as finding "customers who are researching products and actively considering buying a service or product like those you offer" — but as targeting they constrain reach on an awareness buy.
• Demand Gen. Audiences apply at ad-group level, but Google notes that "for YouTube inventory on Demand Gen, these signals are applied as hints" — a target on some inventory, a hint on others.
• Performance Max. In-market can only be added as an audience signal. Google states that signals are optional and that "Performance Max may show ads to relevant audiences outside of your signals if they have a strong likelihood of converting". A signal informs the model; it does not fence it.

Optimised targeting muddies this further, and it is on by default. Google says it "looks beyond manually-selected audience segments", that "your ads may show to users beyond your selected signals", and that it "is automatically turned on for all campaigns" — Display, Demand Gen and eligible Video. So on those types, targeting is a starting point rather than a hard boundary; a hard boundary is a campaign-structure decision, and the same logic governs signals inside Performance Max. None of the pages cited here shows a last-updated date; all read 28 September 2026.

Observation vs targeting: the setting that decides everything

Observation reports on a segment without changing who sees your ads. Targeting restricts delivery to it. Same audience, two completely different purchases — and getting this wrong is the most common in-market mistake we see.

Google's "About 'Targeting' and 'Observation' settings" page puts it directly. On targeting: "you're telling Google Ads who you'd like to reach with your ads or where you'd like your ads to show on the internet". On observation: "the reach of your campaign or ad group isn't affected". For Search and Shopping campaigns the same page states "you can use 'Observation' for Audiences, which includes data segments for Search ads", and lists the recommended usage as Observation.

Observation is not passive. The same documentation confirms you can apply custom bids or bid adjustments to criteria you only observe — you influence the bid without cutting the reach. Note that you cannot add an observation when you first create an ad group; it is a later edit, which is why it gets skipped at setup.

In the interface the choice appears as Narrow targeting or Add observations, per Google's "Select targeting and observation settings" page. Neither page shows a last-updated date; both read 28 September 2026.

Observation first, as a default. Not caution for its own sake: you cannot evaluate Google's inference in your category, market and budget until you have watched it against your own conversions. Targeting first means paying for the experiment and losing the control group at once.

The check we run before switching a segment to targeting

Our own routine, not a Google recommendation, and it carries no measured figure — it is the decision sequence we use on client accounts.

Add the segments that plausibly match the product as observations only. Then read them in this order:

1. Check the conversion count before the rate. If a segment has not produced enough conversions for you to state the number out loud, it has told you nothing yet.
2. Compare the segment against its own ad group, not other segments. The question is whether this slice behaves differently from the average of the same traffic. Cross-segment comparisons quietly compare different query mixes.
3. Make the bid adjustment before the targeting change. If a segment reads better than its ad group, a positive adjustment buys more of it without giving anyone else up. Most accounts skip this step.
4. Only narrow when a segment is strong and stable. Strong once is a month; stable is the same direction across several reads. Membership is a recent signal, so one good month is the weakest possible case for a permanent restriction.
5. Write the reason next to the change. "Narrowed because both segments beat ad-group CPA across three reads" is a note you can reverse. "Narrowed" is not.

On most accounts the segment stays an observation with a bid adjustment on it, permanently. That is a good outcome, not a failure to commit.

The India caution worth making

Our judgement, not a Google statement. Google publishes no market-by-market breakdown of category coverage, no accuracy figure and no membership window — and we will not invent one.

The category list is global; the behaviour feeding a category is not evenly spread across markets, languages or devices. A confident English category name can behave very differently on an India-targeted campaign — often because it was shaped around a market where the journey is mostly online, while the Indian equivalent runs through dealers, WhatsApp or a showroom visit Google never sees.

So treat the category name as a hypothesis and your own observation data as the evidence:

• Never pick segments from the name alone. Add the plausible ones as observations and let the account tell you which of Google's categories describes your buyer.
• When no category fits, stop forcing one. On Display, Video and Demand Gen a custom segment built from terms your buyers genuinely search will usually beat a loosely-related in-market category. On Search, your keywords already do that job.

Where to go from here

Setting this up today: add the segments as observations, leave the reach alone, then move bids before settings.

If you inherited an account with these already on targeting, find out whether anyone knows why. That decision is rarely documented, and switching it back to observation is a cheap way to learn what the restriction was costing.

Alongside structure, match types and measurement, this is one small part of a bigger job — the job our Google Ads management engagements cover. Paid-media management in India runs roughly ₹25,000–₹1,00,000 a month depending on account complexity; fees exclude ad spend.