How to Get Recommended by ChatGPT and Google AI in 2026

A mother in Phoenix types a question into ChatGPT at 1 a.m.: "My son needs help with fentanyl. What are the best treatment centers in Arizona that take Blue Cross?" She gets a confident, specific answer with four named facilities. If your center is not one of them, you did not lose a click. You lost an admission you never knew existed.

The same thing is happening in every service category. A founder asks Claude for "the best web agency for healthcare brands." A homeowner asks Perplexity who to call about a foundation crack. A CFO asks Google's AI Mode which bookkeeping firm to shortlist. The AI answers with names, and those names win the customer before a traditional search ever happens.

Most businesses have no idea whether they appear in these answers, and the data says most do not. This guide explains exactly how ChatGPT, Perplexity, and Google's AI features choose which businesses to recommend, and gives you a concrete 7-step plan to become one of them. We will use addiction treatment centers as the running example because it is the highest-stakes version of this problem, but every tactic applies to any US service business.

The 30-second answer

To get recommended by ChatGPT, Perplexity, and Google AI, you need four things: a consistent, unambiguous business identity across the web (same name, address, services, and positioning everywhere), answer-shaped content on your site that states facts directly with statistics and sources, strong third-party proof (reviews, directories, press mentions, and "best of" list inclusions, which AI trusts more than your own site), and a technically clean, server-rendered website indexed in both Google and Bing. There is no paid placement. Visibility is earned through entity clarity and third-party trust.

Why AI search changes how customers find you

AI search changes discovery because the answer now happens before the click, and most of the time the click never comes. ChatGPT crossed 900 million weekly active users in February 2026, double its February 2025 figure, and 49% of US adults now use AI chatbots, up from 33% in 2024. About 37% of consumers say they now begin searches with AI tools, and 42% prefer a chatbot over a search engine for multi-step research, which is exactly what hiring a service business is.

On the Google side, the change is just as sharp. A randomized field study found organic clicks drop 38% on queries that trigger an AI Overview, Search Engine Land reported a 61% organic CTR decline, and roughly 60-70% of Google searches now end with zero clicks. Small businesses feel this hardest: sites with 1,000-10,000 daily pageviews saw search referrals fall about 60% over two years per Chartbeat data reported by Axios.

Here is the part that should get your attention: the demand side is enormous and the supply side is nearly empty. Around 45% of consumers now ask AI assistants to find local businesses, yet only 1.2% of business locations ever get recommended by ChatGPT for local service queries. Almost nobody has optimized for this. The businesses that move first inherit a discovery channel with almost no competition.

And the visitors who do arrive from AI are better customers. Semrush's analysis values the average AI search visitor at 4.4x a traditional organic visitor, ChatGPT ecommerce traffic converts 31% higher than non-branded organic, and brands cited inside AI Overviews earned 35% more organic clicks than uncited brands. The honest caveat: AI referrals are still only around 1% of most sites' traffic, though that slice grew 206% year over year. Treat AI visibility as the fastest-growing layer on top of SEO, not a replacement for it. Our guide to SEO for service businesses covers the foundation this builds on.

How AI engines actually pick businesses

AI engines pick businesses by cross-referencing what the wider web says about you: your entity data, your reviews, the third-party pages that mention you, and how extractable your own content is. Each engine has different plumbing, and knowing the plumbing tells you where to invest.

ChatGPT: Bing is the gatekeeper

ChatGPT's live web answers are retrieved through Bing's index. A Seer Interactive study found that 87% of SearchGPT citations matched Bing's top results for the same query. The practical consequence is blunt: a site that is invisible in Bing is invisible in ChatGPT's web answers, no matter how well it ranks on Google. For recommendations drawn from training data rather than live search, ChatGPT leans on high-trust sources it saw repeatedly: Wikipedia, major directories, review platforms, and editorial "best of" lists.

Perplexity: the most winnable engine

Perplexity runs a real-time retrieval pipeline that pulls roughly 5-10 pages per query and cites 3-4. It rewards pages that state the answer in the opening lines, contain verifiable specifics rather than vague marketing, and are recently updated: a large share of its citations come from current-year content. Because every Perplexity answer carries citations and its re-ranking favors relevance over raw domain authority, it is the engine where a well-structured page from a smaller site has the best odds.

Google AI Overviews and AI Mode: normal SEO, looser rankings

Google's official documentation says AI Overviews and AI Mode source from the standard Google index with no special markup, files, or AI-specific requirements. Standard SEO is the on-ramp. But the link between ranking and citation is loosening fast: Ahrefs data shows the share of AI citations coming from top-10 ranking pages fell from about 76% in mid-2025 to roughly 38% in early 2026. That kills the comfortable assumption that "we already do SEO, so we are fine." It also means well-structured pages outside the top 10 now have a citation path that never existed in the blue-links era.

The signals all three share

Across engines, the strongest quantitative evidence points off-site. Ahrefs' 75,000-brand correlation study found branded web mentions correlate with AI visibility at 0.664 versus 0.218 for backlinks, with YouTube mentions the single strongest signal at roughly 0.737. Semrush's 150,000-citation analysis found Reddit cited in about 40% of LLM answers and Wikipedia in about 26% (treat the exact shares as directional; they shift month to month). And the only peer-reviewed controlled experiment in this field, the Princeton and Georgia Tech GEO study, found that adding statistics, quotations, and source citations to a page boosts AI visibility 30-40%, while promotional tone underperforms. Being talked about matters more than what you say about yourself, and when you do speak, evidence beats adjectives.

The 7-step fix

The fix is a sequence, not a menu: make your business unambiguous to machines, make your content quotable, then make the rest of the web vouch for you. Work the steps in order. Each one compounds the next.

Step 1: Lock down your entity and NAP consistency

AI models recommend businesses they are confident about, and confidence comes from agreement across sources. Your name, address, phone, service list, and one-sentence positioning must be identical across your website, Google Business Profile, Bing Places, Yelp, BBB, industry directories, LinkedIn, and every other public profile. Contradictory data across listings lowers the model's confidence and quietly removes you from consideration. Claim Bing Places even if you have never thought about Bing: it feeds the engine 900 million people use weekly.

Step 2: Publish answer-shaped content

Restructure your key pages so a machine can lift a complete answer from them. Lead every section with a direct 2-3 sentence answer under a question-form heading, then elaborate. Add the three elements the Princeton GEO experiment proved move citations: statistics, quotations from credible sources, and linked references (statistics addition scored a 41% gain on the study's position-adjusted visibility metric). Write like a reference document, not a brochure. Analyses of ChatGPT's citation behavior show it favors pages whose claims are self-contained and easy to extract verbatim.

Step 3: Add structured data that describes the entity

Implement JSON-LD schema for Organization, LocalBusiness (with hours, geo, service area), FAQPage on question-heavy pages, and Article on editorial content. Be honest about what this does: Google says no special schema is required for AI Overviews, but Microsoft has stated that schema helps its LLMs understand content, which matters because Bing feeds ChatGPT. Schema is entity disambiguation for machines: it tells every crawler exactly who you are, what you do, and where. It is a confidence multiplier, not a magic switch.

Step 4: Build third-party proof where AI already looks

This is the highest-leverage step and the one most businesses skip. AI engines lean on a small set of trusted surfaces: review platforms, Reddit, Wikipedia, YouTube, LinkedIn, and editorial sites. Review platforms function as AI's trust layer: engines cross-check multiple review sites before recommending a vendor, so maintain complete, actively reviewed profiles on the 2-3 platforms that matter in your category. Pursue inclusion in the "best X in [city]" articles that already rank in your niche, because recently updated list posts are the most-cited page format in ChatGPT sources. Pitch podcasts and local press; earned media is the delivery mechanism for the branded mentions that correlate with AI visibility three times more strongly than backlinks. Participate genuinely in relevant subreddits, with real answers under a transparent account. Expect this step to compound over 12-18 months, not 30 days.

Step 5: Build direct-answer pages for your money queries

List the 10-20 questions a buyer asks an AI right before choosing a provider: "best [service] in [city]," "how much does [service] cost in 2026," "[option A] vs [option B]," "does [service] take [insurance/payment type]." Build one page per question that answers it completely, includes the current year in the title, gives concrete numbers, and compares options fairly, including ones you do not sell. Yes, publish your own comparison lists: Ahrefs found vendor-published "best X" lists among the most prominent ChatGPT sources, provided each entry has distinct, concrete rationale. Then keep these pages fresh with substantive updates: AI-cited content is measurably fresher than organic-ranked content (about 25.7% fresher in Ahrefs' 17-million-citation analysis), and much of that freshness comes from maintained older pages, not new ones.

Step 6: Fix the technical foundation

Three checks, in order. First, verify your site in Bing Webmaster Tools, submit your sitemap, and confirm your key pages are indexed; this is a hard prerequisite for ChatGPT citations. Second, confirm your content exists in the initial HTML response. The Vercel and MERJ dataset of 500 million GPTBot fetches found zero evidence of JavaScript execution by AI crawlers: anything your site renders client-side simply does not exist for ChatGPT, Claude, or Perplexity. If your site is a client-rendered React or Vue app, move to server-side rendering or static generation; this is one of the reasons we build on Next.js rather than page builders, a tradeoff we break down in WordPress vs Next.js in 2026. Third, check robots.txt and your CDN or firewall logs to confirm the search-time agents (OAI-SearchBot, ChatGPT-User, Claude-SearchBot, PerplexityBot) are not being blocked. Cloudflare bot rules silently blocking AI agents is the most common invisible killer we find in audits.

Step 7: Measure your AI visibility monthly

Only 14% of marketers actually track AI citation visibility, even though 43% call AI optimization an active 2026 strategy. Tracking is simple to start: run your 10-20 money queries through ChatGPT, Perplexity, and Google AI Mode each month, log who gets named, and note the sources each engine cites. Those cited sources are your target list for Step 4. Watch AI referrals in your analytics (chatgpt.com, perplexity.ai referrers) and treat them as a conversion cohort, not a traffic line: their value shows up in lead quality.

AI visibility for recovery centers: the highest-stakes version

For addiction treatment centers, AI visibility is not a marketing curiosity; it is where families now start. One in three US adults has used AI chatbots for health information and 16% have used them for mental health questions, per KFF polling. OpenAI reports 40 million people ask ChatGPT health questions daily, with over 40% of health users researching treatment options and roughly 70% of those conversations happening outside clinical hours. That is the 1 a.m. mother in Phoenix, at national scale.

The economics make organic AI visibility unusually valuable in this niche. Paid search clicks run $25-$150 with "alcohol rehab near me" reaching about $185 per click, and a residential admission typically costs $4,000-$8,000 to acquire through PPC. Meanwhile, per the 2023 NSDUH, only 15.6% of the 48.5 million Americans with a substance use disorder received treatment. Demand is not the constraint. Findability and trust are, across roughly 17,561 competing facilities.

The playbook above applies, plus four niche-specific moves. First, get every trust credential machine-readable: Joint Commission or CARF accreditation stated in plain text and schema on your site, because these are the signals families, payers, and now AI models use to separate legitimate programs from bad actors. Second, claim and complete your FindTreatment.gov listing. It is SAMHSA's official locator and the government-endorsed starting point that other agencies route help-seekers to. Third, maintain LegitScript certification. It has been mandatory for treatment ads on Google, Meta, and Microsoft since 2018, takes 4-8 weeks, and doubles as a third-party legitimacy signal. Fourth, make your insurance-verification page a first-class asset: industry benchmarks report that around 70% of admissions stall on unclear costs, and "Verify Insurance" CTAs outperform generic CTAs by 32%. "Which centers take my insurance" is precisely what families ask ChatGPT, so publish clear, current payer information in crawlable HTML. We cover the conversion side in depth in our guide to addiction recovery website design that converts.

One ethical line, stated plainly: everything above is about making true information findable. In behavioral health, exaggerated claims, fake reviews, or gaming tactics are not just brand risks, they can harm people at their most vulnerable and violate the compliance regime (LegitScript, state regulators, FTC) that governs this industry. Accuracy is the strategy.

What not to do

The fastest way to waste an AI visibility budget is to buy tactics that sound technical but do nothing. These four come up in almost every audit we run.

Do not treat llms.txt as a ranking tactic

Despite tens of thousands of sites publishing llms.txt files, no major AI engine has confirmed using them. Google's John Mueller said "no AI system currently uses llms.txt", and Ahrefs' log analysis found about 97% of published llms.txt files receive zero AI-bot requests. Adding one is harmless, cheap future-proofing. Anyone selling it as the fix is overclaiming.

Do not mass-produce AI-generated content

Flooding your blog with thin AI-written posts optimizes for a 2019 version of search. The Princeton study showed evidence density wins and promotional, keyword-stuffed content underperforms. Ten pages with real statistics, sourced claims, and specific local knowledge will beat two hundred generic ones.

Do not expect schema alone to do anything

Structured data on top of thin content, inconsistent listings, and zero third-party mentions is a label on an empty box. Schema clarifies an entity that already has substance; it cannot create trust that does not exist elsewhere on the web.

Do not astroturf Reddit or buy reviews

Marketers are actively gaming Reddit because its visibility correlates with LLM mentions, and platforms crack down on exactly this behavior. Fake reviews and sockpuppet threads get deleted, get you banned, and in regulated industries like treatment, invite FTC and regulator attention. Earn the mentions; the compounding is worth it.

Your 30-day action plan

You can complete the foundation in 30 days with a few focused hours per week. Here is the sequence we run for clients.

Week 1: Baseline and entity. Run your 10-20 money queries through ChatGPT, Perplexity, and Google AI Mode; screenshot who gets recommended and which sources are cited. Audit every public listing for NAP and positioning consistency, and fix mismatches. Claim Bing Places and verify the site in Bing Webmaster Tools.

Week 2: Technical. Confirm key content is in the raw HTML (view-source test). Check robots.txt and CDN logs for blocked AI agents and unblock the search-time bots. Submit sitemaps to Bing, fix indexing gaps, and add Organization, LocalBusiness, and FAQPage schema.

Week 3: Content. Rewrite your top 5 service pages answer-first: direct answer under each heading, statistics with linked sources, an FAQ section built from real customer questions. Add or update the current year and a substantive refresh to your two most important pages. If you are a treatment center, ship or upgrade the insurance-verification page this week; nothing else on the site matters more.

Week 4: Third-party proof. Launch a systematic review-request flow on your two most relevant platforms. Identify the "best of" lists and directories your baseline research showed AI citing, and start outreach to the top five. Pitch one podcast or local publication. Then put a monthly 30-minute recurring block on the calendar to re-run the baseline queries and log movement.

When to hire help

Hire help when the checklist stalls on execution: the site needs re-platforming to server-rendered HTML, nobody owns content production, or three months of self-serve effort has not moved the baseline. The 30-day plan above is genuinely doable in-house, and if you run it and start showing up for your money queries, you may not need an agency at all.

Where an agency earns its fee is the compounding work: engineering a fast, server-rendered site AI crawlers can actually read, producing evidence-dense content on a schedule, and running the digital PR that builds mentions for 12-18 months. That combination is what we do at KhanWork: conversion-focused sites built on Next.js, plus SEO and AI search optimization for US service businesses, with deep experience in healthcare and recovery brands. For Cornerstone Healing Center, our flagship treatment-center client, that approach lifted website conversions 20%; the full breakdown is in our case studies, and our pricing is public because cost transparency is advice we take ourselves.

Either way, start the baseline this week. Ask ChatGPT, Perplexity, and Google AI the questions your customers ask, and see whether you exist. In a market where 45% of consumers ask AI for recommendations and 1.2% of businesses show up, the gap will not stay open forever. Right now, it is wide open for whoever moves first.