AI recommendations: the visibility setup for your stack
AI recommendations run on data. Feeds, schema and fresh reviews decide if ChatGPT, Perplexity and Google AI Mode name your business. Wired in an afternoon.
This sheet contains partner links. A purchase through one earns Runbook a commission at no additional cost to you. How we make money.
More of your future customers are asking an AI where to buy instead of searching Google, and the AI only names businesses it can find in its data. If you are not in that data, you are invisible to those buyers. By the end of this build you'll be in it: a product feed the assistants can pull, a Google listing they trust, page labels they can read, and fresh reviews. The stack is free feeds, one plugin, and a review habit. Most of it takes an afternoon.
The shift is measurable. Adobe's 2025 survey put 53% of US consumers on record as having researched a purchase with AI, and across the trillion-plus retail visits Adobe measured, AI-referred sessions multiplied through the 2025 holiday window while converting about a third above the other channels. So when the AI names someone, how does it pick, and how do you get named?
What you'll build
A setup that makes the assistants able to find you and willing to quote you:
- A product or business feed the shopping answers actually read.
- A Google Business Profile (your free listing on Google Maps and search) and clean page labels the AI can parse.
- A review habit that gives the AI a reason to trust you.
- A capture layer so the ready buyer it sends does not bounce.
Stack
All three big assistants turned recommendations into a core feature in the last eighteen months. Each reads from structured business data plus trusted third-party sources like reviews.
- ChatGPT added shopping research in 2025 with sourced product details and comparisons. It briefly tested letting people buy inside the chat, then in March 2026 went back to sending buyers to the seller's own website. (CNBC reported the pivot.) Discovery stayed. The sale moved back to you.
- Perplexity built a shopping flow with in-chat checkout and a merchant program, which is a free sign-up where you hand it your product catalog, specs, and reviews. It leans heavily on reviews when it answers.
- Google AI Mode (Google's chat version of search) pulls from a catalog of tens of billions of product listings, and in January 2026 added new Merchant Center attributes and a business-chat agent aimed at conversational questions. (Google's announcement has the detail.)
The build sits on free feeds, a plugin, and a review habit. Two paid tools show up later: Semrush (a search-data tool) shows which questions the assistants pull you into, and Surfer helps you shape the pages they quote.
On the surfaces and in the stack
Perplexitymerchant program for catalogs and reviews
Semrushsee which queries the assistants pull you into
Surferbrief the pages they citeSteps
You do not chase every platform. You make yourself easy to find, easy to read, and well reviewed. In order:
The build, in order
Get into the feeds
Fix your Google Business Profile
Add the basic schema
Make reviews a habit, not a campaign
Build pages the assistants can quote
Your move this week
The part that breaks
The second trap is deciding AI traffic is too small to bother with. Most analytics setups do not label AI visitors as their own channel yet, so owners cannot see them and assume the number is zero. The people who arrive after an AI pointed them your way show up half-sold, which is why they convert better. Build for the channel while it is still quiet.
It is the same visibility problem that decides whether you show up in Google's AI Overviews as a small business, and it pairs with building a name people search for.
Copy this
A review-request message you can drop into your CRM automation, sent the day after a job closes:
Hi [first name], thanks for choosing us.
If we did right by you, a quick Google review helps
more than you'd think: [direct review link]
Takes 30 seconds and it means a lot. Thank you.
One link, no friction. A steady flow of fresh reviews is the signal the assistants weight.
Upgrade path
Once feeds, listing, labels, and reviews are live:
- Route AI-sent and ad-sent leads into one inbox with instant follow-up, because the sale is moving back to your site and a recommendation only counts if it turns into a booking. The wiring lives in sending leads straight to your CRM with Make.
- Get mentioned where the AI reads: an honest comparison listing, a Reddit thread where real people vouch for you. You cannot fake this, which is exactly why the AI trusts it.
This pairs with the capture logic in the Meta Advantage+ build. The writing and SEO hub tracks the shift as it moves.
Frequently asked questions
How do I get my products into the feeds AI assistants read?
Set up a Google Merchant Center feed if you sell products, then enroll in Perplexity's merchant program for catalogs, specs and reviews. The feeds are plumbing, not marketing, and they are the single biggest lever for product recommendations: no feed, no mention.
Which schema types should I add first for AI recommendations?
LocalBusiness, Product, Review and FAQ markup, in that order of relevance to your business. On a mainstream website platform a schema plugin handles most of it, so this step rarely needs a developer. Clean markup is what lets the systems parse and trust your pages.
How long does the AI-recommendation build take?
One afternoon covers the one-time wiring: the feed, a complete Google Business Profile and the schema plugin. The review habit is the ongoing part, an automated ask after every job, and it is also the piece the assistants weight most heavily over time.
How do I check whether AI assistants already send me traffic?
Open the referrer report in your analytics and look for visits from domains like chatgpt.com and perplexity.ai. Most setups do not label these as a channel yet, so owners assume the number is zero. The visitors who do arrive tend to show up half-sold, which is why they convert better.
About Runbook
AI tools and automation builds for marketers. What to use, how to wire it, and the workflow to copy this week. How we work
GET THE NEXT DISPATCH
Run the next build before your competitors read about it.
One short email when an AI tool or automation actually changes the work, with the build to copy.
No send unless there is a build worth running.
// keep_reading
Related builds
Google AI Overviews: the content build that survives it
Google AI Overviews now answer half of searches. Branded search, buying-intent pages and local Maps still send clicks. Wire the build in a week.
The move. Stop building for clicks that are vanishing. Build a system that owns your branded search, ranks buying-intent pages, and wins local on Maps.
Best SEO content tools in 2026
Frase for content briefs, Surfer for on-page scoring, Semrush for keywords. The SEO content tools for 2026, with pricing and where each trips you up.
Branded search monitoring: the one-hour setup
Branded search monitoring in one hour: own the first screen for your name, measure demand for it monthly, and wire the review habit that grows it.
The move. Build demand for your name, not just your category. A branded search you own, a clean first screen, and a review habit are the traffic AI cannot answer for you.