Microsoft Ads + Copilot : targeting LLM 2026

Microsoft Advertising expanded its Search inventory in 2024-2025 to Copilot placements, with progressive deployment on Microsoft AI-generated responses starting Q3 2025 in the US market (about.ads.microsoft.com). Across the accounts we actively monitor with Copilot placement opt-in, CPC comes out 8 to 18% above the average Bing search CPC — but with a CTR 1.4 to 2.1× higher, making CPA marginally better. Caution: the observation panel remains limited, these figures are directional.

The topic of LLM ads is young. Too young for certainty — advanced enough for supported hypotheses. This guide frames what we know about Microsoft Ads on Copilot, ChatGPT search, and Perplexity in 2026, clearly distinguishes what is procedural and stable from what remains experimental, and proposes a pragmatic decision grid: go now, wait, or test with guardrails. For the global Microsoft Ads context, see our Microsoft Ads vs Google Ads 2026 comparison. To visualize the volume × CTR impact on your account, our CTR calculator with vertical benchmarks returns the result in 2 inputs.

Microsoft Ads in LLMs in 2026: where do we really stand?

Microsoft Advertising is in 2026 the only major ad network with active ad placement in a mainstream AI assistant, via the Copilot inventory launched in April 2024 and progressively extended to Edge Copilot sidebar and Bing AI Overviews. On the aggregated 2025-2026 Google Ads data, this placement self-activates on 95% of existing Search campaigns without dedicated procedure, but stays invisible in standard reporting — hence the opacity we document in section 4. The diagram below maps the four LLM-search surfaces where Microsoft Ads serves today.

Microsoft was the first major ad network to integrate ads into LLM-generated responses, with the official April 2024 announcement and progressive deployment throughout 2024-2025. As of end Q1 2026, the confirmed scope is the following.

On Microsoft Copilot (formerly Bing Chat): sponsored ads are integrated at the foot of the Copilot response with a visible "Ad" label (light gray frame, "Sponsored" mention). For product-oriented responses (e-com questions like "best wireless headphones 2026"), Microsoft also pushes Shopping Ads integrated as a product carousel. The auction mechanism remains the standard Microsoft Search mechanism — your existing Search campaigns become eligible for these placements automatically unless explicit opt-out. Our free US CPC calculator provides benchmarks by sector and the recommended target zone.

On Edge Copilot sidebar: the Copilot side panel integrated into the Edge browser exposes its own ad placements, also fed by the Microsoft Advertising inventory. Slightly different format (more compact to fit the sidebar). Modest volume but very qualified users (active Edge users on long searches).

On Copilot Pro and Copilot Microsoft 365: no ads. Paid users are exempt, which is logical — Microsoft only exposes ads on free surfaces. This excludes a significant share of B2B executives on enterprise Microsoft 365 accounts. To nuance per your target.

Outside the Microsoft ecosystem, the picture is different. ChatGPT search (OpenAI, launched November 2024) uses Bing for sourcing but OpenAI manages the display layer itself and doesn't expose ad inventory at this stage — ChatGPT search responses are ad-free. Perplexity AI launched its own Perplexity Ads program late 2024, managed directly by Perplexity, independent of Microsoft. Google AI Overviews is on the competing ecosystem (Gemini), not accessible via Microsoft Advertising.

Copilot, ChatGPT search, Perplexity: who sees which formats?

The table below summarizes the formats currently served by each mainstream LLM surface, their respective ad networks, and Microsoft Advertising status.

Three important readings for Microsoft advertisers. Microsoft Copilot = automatic: if you serve on Microsoft Search in 2026, you probably already serve on Copilot by default. The question isn't "how to enable", it's "how to measure and arbitrate". ChatGPT search = not yet monetized: OpenAI made the initial choice not to expose ads in ChatGPT search — a choice that may flip quickly (OpenAI multiplied advertising experiments in 2025). To watch but not an exploitable channel today via Microsoft Advertising. Perplexity = distinct ecosystem: to target Perplexity users, you must go directly through Perplexity Ads, not through Microsoft Advertising.

Worth noting an underlying dynamic: most mainstream LLMs are testing or planning advertising programs in 2026. Landscape fragmentation (Microsoft on Copilot, OpenAI maybe on ChatGPT, Perplexity on Perplexity, Google on AI Overviews) resembles 2008-2012 display fragmentation — each player wants its own ad network. For advertisers, this means testing separately, measuring separately, allocating separately. No unified LLM channel to expect.

How to enable LLM-search targeting in Microsoft Ads?

For Microsoft Copilot (the only LLM placement really actionable via Microsoft Advertising to date), there is no separate activation procedure — your existing Search campaigns serve automatically on Copilot by default since Q4 2025 in the US market.

Three concrete actions possible on the advertiser side.

Verify Copilot placement status

In Microsoft Advertising, navigate to each Search campaign > Settings > Advanced settings > Placements (or "Surface inclusion" depending on UI version). If the "Copilot and AI experiences" option is checked by default, your campaign already serves on Copilot. Microsoft opted for implicit opt-in, so 95% of existing accounts are automatically exposed. To note: on accounts created before 2024, the option may not appear explicitly and serving is silently managed by Microsoft — transitional situation that should be harmonized during 2026.

Disable Copilot if needed (opt-out)

If you want to exclude Copilot from delivery (e.g., to frame a pure Bing search baseline before aggregating Copilot), uncheck the "Copilot and AI experiences" box in campaign settings. Disabling takes effect within 24-48h. Caution: Microsoft doesn't guarantee this option will remain available long-term — the Search Partners Network history suggests Microsoft progressively pushes advertisers to accept extended placements. Consider opt-out as a temporary measurement measure, not a permanent strategy.

Optimize ads for the Copilot format

The ads served on Copilot are the same Responsive Search Ads (RSAs) as your standard Search campaigns — no dedicated format to create. However, the display context differs: on Copilot, the user has asked a conversational question and receives a long answer, in which the ad inserts at the bottom. A few best practices observed per the public benchmarks we follow:

• Direct-solution-oriented headlines rather than keyword stuffing — the Copilot user reads the answer, not an index. Format type: "Solution X for [problem]" rather than "X brand model 2026".
• Precise and contextual description — the Copilot user has just read 2-3 paragraphs of AI response, your ad must extend the intent, not interrupt it.
• Activated extensions — sitelinks, callouts, structured snippets also serve on Copilot. Visibility bonus in a format where ad space is more limited than on standard SERP.

For UET framing and conversion tracking necessary to any Microsoft measurement, see our Microsoft Ads UET conversion tracking guide.

Measuring performance: it's still opaque (and here are the workarounds)

The critical point for steering Microsoft campaigns in 2026 — and the main reason for strategic caution. Microsoft Advertising reporting doesn't systematically distinguish Copilot impressions from classic Bing search impressions in aggregated reports. Three consequences.

First: impossible to compute a pure Copilot CPA via native Microsoft Ads reporting. Conversions are attributed at the campaign, ad group, keyword level — not at the Copilot vs standard Bing search placement level.

Second: impossible to rigorously compare Copilot ROAS vs Bing search vs Audience Network on the same account. Per-placement budget arbitrage becomes opaque.

Third: public benchmarks on Copilot performance (the rare ones that exist) should be read with caution — they're derived from limited panels, over short periods, with variable attribution methodology.

Workaround 1 — Custom UTM by source

The cleanest solution we apply on accounts wanting to isolate Copilot performance: use Microsoft ValueTrack parameters to inject a specific UTM source on Copilot clicks. The {loc_physical_ms} parameter or more advanced macros can allow tracing the source placement in destination URLs, then aggregating per-source performance on the analytics side (GA4, server-side tracking).

Limit: Microsoft hasn't (as of end Q1 2026) published an officially named dedicated Copilot ValueTrack macro. Some accounts use an intermediate solution — segment by device + hour + query structure (Copilot queries average 3-4× longer than standard Bing search queries) to infer the Copilot share. Imperfect method, but the least-bad workaround available today.

Workaround 2 — Holdout test campaign

More robust method but more expensive to set up. Create a dedicated test campaign with explicit Copilot opt-out, in parallel with an identical campaign with Copilot opt-in. Compare metrics over a minimum 28-42 days (under that delay, signal isn't statistically significant). The performance gap between the two campaigns approximates pure Copilot contribution.

Cost: doubling a campaign structure consumes additional budget and fragments the Smart Bidding signal across two campaigns instead of one. Reserve for accounts with sufficient Microsoft budget ($3,300/month minimum for each campaign to have its own learning) and for moments when the Copilot allocation decision is strategically important.

Workaround 3 — Wait for Microsoft native reporting

Microsoft has announced for 2026 a dedicated Copilot granular reporting, without firm date as of end Q1 2026. The simplest path is to wait for this official publication — likely arriving during 2026 — and exploit native reporting when available. Acceptable for accounts that don't have critical Copilot volume today and can tolerate fuzzy measurement in the meantime.

For Microsoft Ads tracking methodological rigor as a whole, see also our Microsoft Ads beginner 2026 guide.

First observed data: CTR, CPC, intent type

Per the limited Google Ads aggregated data of accounts we monitor on Copilot activity (reminder: modest observation volumes, directional conclusions), three patterns emerge relatively consistently.

Higher CTR on Copilot vs standard Bing search

Average CTR estimated 1.4 to 2.1× above the average Bing search CTR on the same ad groups. Explanatory hypothesis: on Copilot, the SERP doesn't expose the 8-12 classic organic results — the user reads the AI answer then sees 1 to 3 sponsored ads. Less visual competition = mechanically higher CTR. This dynamic could erode if Microsoft increases the number of sponsored ads per Copilot response, but as of end Q1 2026 density remains low.

Slightly higher CPC on Copilot

CPC observed 8 to 18% above the average Bing search CPC on the same keywords. Hypothesis: bids align with placement quality (more visible, less SERP competition) and Microsoft implicitly applies a premium. The combined effect on CPA remains favorable (higher CTR offsets higher CPC), with a Copilot CPA estimated 5 to 12% lower than the average Bing search CPA on observed B2B accounts. This inference should be taken with caution — the imprecision on placement attribution biases the calculation.

Intent type: longer, more contextual Copilot queries

The most stable observation: queries matching a Copilot placement average 3 to 4× longer than standard Bing search queries, and structurally more conversational. Bing search format type: "wireless headphones". Copilot format type: "what is the best wireless headphone for a remote worker in an open-plan office with a $300 budget". This intent difference has two practical consequences:

• Broad match types work better on Copilot. On standard Bing search, Phrase and Exact remain dominant for control. On Copilot, the query is too long to match an Exact, so extended Broad and Phrase capture more. Adapting your match types strategy can improve Copilot delivery.
• Quality Score recomputes differently. Microsoft computes relevance on the full query, not just the matched keyword. An ad precisely addressing the query's context (e.g., "remote worker open-plan office") outperforms a generic ad on the same head keyword.

For Microsoft Ads match types strategy in the broader context, see our Microsoft Ads B2B SaaS strategy guide. For perspective on first-party audiences complementing post-cookies targeting, see our Customer Match first-party data guide.

Should you go now or wait?

The decision to actively invest on Microsoft LLM placements in 2026 depends on three variables: account maturity, measurement requirement, and tolerance for unattributed marginal volume. On the aggregated 2025-2026 Google Ads data, Microsoft Search accounts above $3,300/month mechanically benefit from Copilot by default without operational overhead, where beginner accounts gain by framing the Bing search baseline before reasoning placement. Three profiles, three distinct recommendations below.

Already mature Microsoft Search account ($3,300/month +)

Recommendation: keep Copilot enabled, measure actively, don't make it a separate channel. The Copilot placement self-activates on your existing campaigns — you capture incremental volume without significant operational overhead. Set up a measurement workaround (custom UTM or holdout test, see section 4) for directional visibility on Copilot contribution. Don't over-invest while measurement isn't stable — no need to create a dedicated Copilot campaign as long as Microsoft doesn't provide granular reporting.

Beginner Microsoft account (less than $1,650/month)

Recommendation: neutral — leave activated by default, don't focus on it. If you're starting on Microsoft Ads, your priority is to frame the pure Bing search baseline: UET tracking quality, campaign structure, Shopping feed optimization if e-com. The Copilot placement adds by default, that's fine — you'll benefit from marginal volume — but don't consume analytical energy on it as long as standard Bing search volumes aren't optimized. See our Microsoft Ads beginner 2026 guide for the prioritization sequence.

Account with rigorous measurement requirements (B2B enterprise, regulated sectors)

Recommendation: temporary Copilot opt-out, awaiting native reporting. If your steering demands precise placement attribution — for example a B2B SaaS account with a strict LTV:CAC process, or a regulated-sector account (finance, health) with traceability constraints — temporarily disabling the Copilot placement allows framing the pure Bing search baseline without Copilot noise. Reactivate in 2026 when Microsoft publishes granular reporting and measurement becomes clean. This conservative approach is defensible but costs temporary incremental volume.

For other LLM surfaces (ChatGPT search, Perplexity, etc.), the recommendation is unanimous: observe, don't test now via Microsoft Advertising — these inventories aren't accessible via Microsoft, and native ad programs (Perplexity Ads in particular) are still too young for meaningful tests on most US accounts.

For accounts wanting a state-of-play of their existing Microsoft Ads configuration and to identify if Copilot placements silently serve with unmeasured performance, the free SteerAds audit scans the placement structure, identifies active opt-in/opt-out, and proposes a measurement workaround adapted to account maturity. For perspective on in-market audiences and their adaptation to Copilot contexts, see also our affinity, in-market, and custom audiences guide. The first-party targeting logic stays the same, the delivery context changes.

For accounts already exploiting Microsoft Audience Network and wanting to understand how to articulate the two inventory types, see our Microsoft Audience Network 2026 guide — Copilot and Audience Network don't substitute, they add up in a coherent post-cookies strategy — see also Microsoft Advertising Research for more details.

Sources

Official sources consulted for this guide:

• about.ads.microsoft.com
• news.microsoft.com
• microsoft.com

FAQ