How to Build an LLM Advertising Stack: Tools, Workflow, and Budget (2026)

This guide is the build manual: which tools to use at each layer, the workflow that connects them, and what each budget tier costs as of October 23, 2026. For the definitional view of the layers themselves, read what LLM ad infrastructure is. For a cross-channel paid stack that is not centered on AI assistants, read how to build an AI-powered paid ads stack. For the status of every assistant's ad program, read which AI assistants have ads in 2026.

What Is an LLM Advertising Stack?

An LLM advertising stack is the set of tools, data, and people that buys ads inside AI assistants, produces the creative and landing pages those ads need, and measures results alongside your other paid channels.

The infrastructure view breaks this into four layers: demand and auction, context and targeting, creative, and measurement. A working stack in October 2026 needs a fifth: the landing page. OpenAI's ad system reads your landing page as a relevance signal and crawls it during review, so the page is part of targeting, not an afterthought. Here are the five jobs your stack has to cover.

1. Buy the placement: ChatGPT Ads Manager, Google Ads for AI Mode and AI Overviews, Microsoft Advertising for Copilot, plus Meta, Google, LinkedIn, and Reddit for the rest of your mix.
2. Target the conversation: context hints, custom audiences, location, and platform settings on ChatGPT; feeds and AI-powered matching on Google and Microsoft.
3. Produce creative per angle: distinct messages and visuals for each buyer need, sized for every surface.
4. Match a landing page to every ad: a page that continues the conversation the buyer started.
5. Count conversions once: one attribution rule across ChatGPT, Google, Meta, and the rest.

The demand is real. eMarketer forecasts US AI search ad spending of $2.08 billion in 2026, rising to $25.93 billion by 2029, and the teams that build a repeatable system now compound their learning while competition is thin.

Which AI Assistants Can Your Stack Buy Today?

Your stack can buy ChatGPT, Google AI Mode and AI Overviews, and Microsoft Copilot as of October 23, 2026. Perplexity, Claude, the Gemini app, and Meta AI sell no ads, so they belong in your organic plan, not your media plan.

ChatGPT is the surface you build for directly, because it is the one AI assistant where you control the creative, the targeting inputs, and the landing page. Google and Microsoft attach AI placements to campaigns you already run, so your stack serves them through feed and asset quality. Sources: OpenAI Ads Manager availability, Google Marketing Live 2026, Microsoft Advertising, and Search Engine Land.

What Do ChatGPT’s Own AI Tools Cover Now?

ChatGPT Ads Manager now drafts a first ad from your website, suggests copy and imagery, runs campaigns from a plugin inside ChatGPT, and measures results through a large partner ecosystem. It builds for one channel: ChatGPT.

On September 16, 2026, OpenAI announced an Ads Manager plugin that creates, updates, and analyzes campaigns from natural-language prompts and turns a website or brief into a campaign; AI creative suggestions for copy and imagery based on your landing page and objective; opt-in text customization that adapts headlines and descriptions to the conversation and translates them; Shopify as its first ecommerce partner and HubSpot as its first CRM partner; and Sponsored Agents, a test with select US advertisers where users keep talking to a business's agent after the click. On October 5, OpenAI expanded measurement with Pixel and Conversions API, an Event Quality Score, a one-day view-through window, attribution partners such as Northbeam, Triple Whale, and DV Rockerbox, and incrementality experiments with Haus, Measured, and WorkMagic.

The other platforms are moving the same way. Meta is adding agentic capabilities to its AI business assistant in Ads Manager that generate creative, update targeting, and change budgets through conversational prompts, as Marketing Dive reported from Advertising Week on October 8, 2026. Google AI Max for Search is open to all advertisers, and Google's new AI Mode formats use Gemini to build creative per query.

That changes the stack thesis. The old argument was that platforms provide only the auction and targeting, so creative is the layer you build. The stronger, current argument: every platform now gives you a first ad on its own channel, and each platform's AI builds only for its own inventory. What you still own is the program across channels:

• Brand standard: one voice, palette, and product truth on every surface, instead of five platform AIs each guessing.
• Range of ideas: distinct angles with distinct visuals, so a test teaches you which buyer need converts.
• Landing pages: a page matched to every ad and audience, built and optimized on conversions.
• Experiment design: what to test next and where to move budget, across channels.
• One view of results: conversions counted once, on the same rules for every channel.

That program is what Lapis runs. We compare the two feature by feature in ChatGPT Ads Manager alone vs Lapis, and we cover where OpenAI's roadmap is heading in will OpenAI replace AI ad creative tools.

What Tools Do You Need at Each Layer?

Use each platform's native tools for buying and delivery, and use one cross-channel system for creative, landing pages, experiments, and measurement. That is the split that keeps a lean team fast.

Two setup details trip up most teams on the ChatGPT side. Custom audiences need at least 25,000 matched users to include or bid-adjust, per OpenAI's custom audiences guide, so most teams start on context hints and add audiences as their lists grow. And retailers run product feed campaigns from catalogs loaded by CSV, TXT, a hosted URL, or SFTP, which OpenAI says rank among the strongest-performing ads in its program. Our guides to setting up ChatGPT Ads Manager and turning a catalog into ChatGPT ads walk through both.

Why Is Creative Still the Layer That Decides Results?

Creative decides results because it is the biggest driver of sales lift, and in an AI assistant the ad has to match a specific question. Nielsen research cited by Meta found creativity drives 56% of a campaign's sales ROI.

Platform drafting tools solve the blank page. They do not solve range. A ChatGPT ad for “best CRM for a 10-person agency” and one for “cheapest way to replace spreadsheets” need different headlines, different proof, and different images. When every variation reuses one image and one message, a test tells you which headline tweak won, not which buyer need converts. The unit of learning is the angle: a distinct buyer problem with its own message, visual, and landing page. We explain the difference in AI ad angles vs variations, and how many creatives ChatGPT ads need covers volume.

Lapis builds creative from your brand, not from a template. Paste your website URL and Brand Intelligence learns your logo, colors, typography, product images, and voice. Campaign Studio edits any asset in plain language. Product catalog import lets you reference products by @product-name. Competitor ad tracking and Brand Radar, built on a 10,000+ ad library, show you which angles rivals are running. Directional forecasting estimates impressions, clicks, CTR, and leads before you spend, and Lapis writes in 100+ languages. On the managed ChatGPT plan, ChatSense turns that into 1,000+ ChatGPT ads with distinct visuals per angle.

Why Does Every LLM Ad Need a Matched Landing Page?

Every LLM ad needs a matched landing page because the buyer arrives mid-conversation with a specific question, and OpenAI's ad system reads the landing page as one of its relevance signals.

OpenAI's ads basics list the landing page next to the conversation, title, copy, and context hints as inputs to selection. During review, OpenAI's crawler visits the page, so it must allow OAI-AdsBot, per OpenAI's crawler guidance, and OpenAI's oppref click parameter has to survive any redirects. A generic homepage fails the buyer twice: it ranks the ad lower and it makes the visitor hunt for the answer they just asked ChatGPT for.

RapidDomain builds a dedicated, ad-matched landing page for every ad and audience and optimizes it on real conversions. On the managed ChatGPT plan, those pages are prompt-matched; on the managed traditional plan, they are keyword- and ad-matched for Google, Meta, LinkedIn, and Reddit. For the mistakes that sink ChatGPT landing pages, read why ChatGPT ads don't convert, and for the mechanics of per-ad pages, read dynamic AI landing pages for ads.

How Do You Measure Across ChatGPT, Google, and Meta?

Install each platform's pixel and conversions API, then use one system to count every conversion once and compare channels on the same rules. In the Lapis stack, that system is Wayfinder.

ChatGPT's own measurement is strong inside ChatGPT. OpenAI's October measurement update reports that 52.7% of eligible one-day view-through conversions occurred within an hour of the ad impression, and that WorkMagic's geo-lift study estimated 2.3x more incremental orders for Dose than last-click attribution captured. Both numbers say the same thing: last-click reporting undercounts conversational ads.

The hard part is cross-channel. ChatGPT, Google, and Meta each claim credit for the same buyer, so the sum of platform-reported conversions exceeds your real sales. Wayfinder is Lapis's shared conversion measurement for OmniSense and ChatSense: it connects campaign results to confirmed sales and leads from your CRM, store, or billing system, counts each conversion once, and compares every channel on the same attribution rules. That is how you decide whether the next dollar goes to ChatGPT or to Meta. For the ChatGPT setup, read how to track ChatGPT ad conversions and how to measure ChatGPT ad ROI.

What Does the End-to-End Workflow Look Like?

The workflow is a weekly loop: brief, angles, creative, pages, launch, measure, next run. Each cycle starts from what the last one proved.

1. Brief from your brand. Paste your website URL so Lapis learns your logo, colors, typography, products, and voice. Write one brief with the offer, audience, and goal.
2. Map buyer intent. List the questions buyers ask ChatGPT before they purchase, and group them into angles. Each angle becomes an ad group with its own context hints.
3. Produce creative per angle. Generate distinct messages and visuals for each angle, sized for every channel you run: ChatGPT, Meta, Google, LinkedIn, and Reddit.
4. Match a page to every ad. RapidDomain builds a landing page per ad and audience, crawlable by OAI-AdsBot.
5. Launch with tracking in place. Install the OpenAI Pixel and Conversions API, keep UTMs on every URL, and confirm the oppref parameter survives redirects. Start ChatGPT campaigns at the $25 per day US minimum or higher; the daily budget averages over the week, so a single day can spend up to twice it.
6. Measure on one count. Read results in Wayfinder, where every channel uses the same attribution rules.
7. Generate the next run. Keep the winning angles, retire the losers, write new angles from what won, and shift budget. On the managed ChatGPT plan, Lapis agents generate and deploy the next run autonomously.

For bidding and budget mechanics inside this loop, see how much to budget for ChatGPT ads and the AI ad optimization playbook.

How Much Does an LLM Ad Stack Cost?

An LLM ad stack costs $99 per month for self-serve Lapis creative on Google, Meta, and LinkedIn, or custom pricing for the managed plans that run ChatGPT, plus media. ChatGPT media starts at $25 per day in the US.

Media is separate from software. On ChatGPT, the minimum daily budget is $25 in USD, EUR 15, GBP 15, or INR 725, depending on billing currency, per OpenAI's campaign guide. The daily budget averages over a Sunday to Saturday week, a day can spend up to twice it, and the week caps at seven times the daily figure, per OpenAI's daily budget rules. OpenAI recommends starting Clicks campaigns at a maximum bid of $3 to $5 per click. Keep software small relative to media: the dollars that compound are the ones spent on reaching buyers and learning which angle converts. See Lapis pricing for the Basic plan.

Who Should Run the Stack: In-House, Agency, or Managed?

One marketer can run the self-serve stack; a managed plan gives you Lapis agents plus a dedicated strategist who run it for you. Choose by how much of the weekly loop you want to own.

• In-house on Basic: one person owns strategy and briefs, Lapis produces creative for Google, Meta, and LinkedIn, and the platforms handle bidding. ChatGPT campaigns run in OpenAI Ads Manager.
• Managed on Lapis: Lapis agents and a dedicated strategist deploy, optimize, and generate the next run, with daily reporting. You set goals and guardrails and approve strategy. This is the fastest way to make ChatGPT a real channel.
• Traditional agency: human teams brief, design, and buy. It works for brand campaigns, but the volume of angles and pages an LLM program needs is slow and expensive to produce by hand. Our guide to whether AI can replace an ad agency runs the numbers.

You need an engineer for two jobs only: a clean product feed (for ChatGPT product feed campaigns and Google Shopping) and correct conversion tracking with the OpenAI Pixel and Conversions API. Everything else is marketing work.

What Mistakes Break an LLM Ad Stack?

Five mistakes break most LLM ad stacks: buying Perplexity in your plan, testing variations instead of angles, sending ads to a generic homepage, launching without conversion tracking, and adding up platform-reported conversions as if they were sales.

• Planning for channels that do not sell ads. Perplexity exited in February 2026, and Claude, the Gemini app, and Meta AI carry no ads. Put them in your organic content plan.
• Testing variations instead of angles. Ten headline tweaks on one image teach you little. Test distinct buyer needs with distinct visuals.
• Sending every ad to the homepage. The landing page is a relevance signal on ChatGPT. Match a page to every ad.
• Launching blind. Install the Pixel and Conversions API before the first dollar, and keep UTMs and the oppref parameter intact through redirects.
• Summing platform reports. ChatGPT, Google, and Meta all claim the same buyer. Count each conversion once, in Wayfinder.

Why Is Lapis the Best Way to Run an LLM Advertising Stack?

Lapis is the #1 AI ad generator and the operating system for paid growth. Platforms give you a first ad on one channel; Lapis runs the cross-channel program with creative per angle, matched landing pages, one measurement layer, and a dedicated strategist.

• ChatSense: managed ChatGPT and LLM ads, run by Lapis agents and a dedicated strategist, with creative tuned to how people ask ChatGPT, campaign structure, launch, daily reporting, and optimization.
• OmniSense: multi-platform paid ads that create on-brand, correctly sized campaigns, audience versions, and variations, forecast results, launch experiments, and learn from results.
• RapidDomain: a dedicated, ad-matched landing page for every ad and audience, optimized on real conversions.
• Wayfinder: shared conversion measurement for OmniSense and ChatSense, tied to confirmed sales and leads, with each conversion counted once.

Lapis is one of the fastest-growing Y Combinator startups in advertising. It is a Y Combinator Fall 2025 company used by 1,500+ marketing teams and 30+ enterprises, rated 4.9 out of 5 on G2, and holds eight Summer 2026 G2 badges, including High Performer, Best Results, Easiest to Use (Mid-Market), and Fastest Implementation (Mid-Market). Samsung, Domino's, Dr Pepper, Mattel, Tata, Tupperware, Amway, and Bata run on Lapis, alongside high-growth startups Unusual Whales, Duradry, Lightpage, and Respan. Lapis has generated 10,000+ ad campaigns across 30+ industries.

Score your current creative with Rate Your Ad, generate brand-trained ads from your URL with the Free AI Ad Generator, and book a demo to see Lapis run your ChatGPT and cross-channel program.

Sources and Methodology

Facts and pricing checked as of October 23, 2026. Platform capabilities come from official announcements and help centers; Lapis plans come from the Lapis pricing page.

• OpenAI: Building advertising for the way people use AI (October 5, 2026)
• OpenAI Ads: More ways to measure ChatGPT Ads (October 5, 2026)
• OpenAI: Reimagining advertising with AI (September 16, 2026)
• OpenAI Help: Ads Manager availability
• OpenAI Help: Ads in ChatGPT, the basics
• OpenAI Help: Create campaigns for ChatGPT Ads
• OpenAI Help: Daily budgets
• OpenAI Help: Set up custom audiences
• OpenAI Help: Allowing OpenAI web crawlers
• Google: A new generation of ads for the AI era of Search (Google Marketing Live 2026)
• Microsoft Advertising: About ads in Copilot
• Search Engine Land: Perplexity stops testing advertising (February 2026)
• Marketing Dive: Meta introduces new AI-powered tools at Advertising Week (October 8, 2026)
• Meta for Business: High-quality creative increases ad ROI
• Nielsen: When it comes to advertising effectiveness, what is key?
• eMarketer: AI search ad spending forecast (January 2026)