The Complete Guide to Getting Your Brand Recommended by ChatGPT

Open ChatGPT tonight and ask it who to buy from in your category. Phrase it the way your buyer would phrase it. If the answer names three competitors and skips the brand you've spent years building, you've found a revenue leak that no SEO report ever showed you.

I run AI and growth at NuVision Auto Glass, a $48M US auto glass company, and my growth company works with US brands in the $5M to $100M range. The question founders at this scale ask me has changed. Two years ago they asked how to rank on Google. Now they ask why ChatGPT recommends a competitor by name and leaves them out.

The traffic numbers explain why this climbed the agenda. AI referrals, meaning visitors who click through to a site from an AI tool's answer, account for about 1% of all web traffic. That number grew about 340% year over year according to the 2026 Conductor benchmark. ChatGPT holds about 87% of it, with Perplexity, Claude, and Gemini splitting most of the rest.

Two other findings matter more than the volume. People arriving from an AI recommendation stay about 2.3 times longer than people arriving from a normal search. And a brand named in a ChatGPT answer is about 2.5 times more likely to get a site visit within 7 days.

So the channel is small, the intent is high, and it compounds fast. At $5M in revenue that's interesting. At $50M, with a decade of brand equity on the line and buyers who now start their research inside a chat window, it belongs on the board agenda next to pricing and hiring.

How the model decides which brand to name

Picture asking a well-read industry peer which firm to hire. They've never bought from anyone in your space. They go off everything they've read: trade publications, review platforms, directories, forum threads, a comparison article from two years ago, and each vendor's own website. They name whoever came up most often, in the most trusted places, with the clearest description attached.

That's the process. Google ranks pages. A language model, the system behind ChatGPT, summarizes reputation. It assembles an answer from everything it has read about your category, and increasingly from a live search it runs at the moment someone asks.

Which leads to the finding that should reorder your marketing budget. About 84% of the citations AI tools show, the source links under an answer, come from earned media, meaning places other than your own website. Directories, review platforms, publications, forums, YouTube.

Now hold that against your current spend. Most brands at this scale put the bulk of their digital budget into their own site and their own ads, the two places the model weighs least. Your website still matters, and I'll cover it first because it's the fastest fix. The deciding mostly happens off your property, on pages you don't own and your current reports don't track.

Layer 1: your website, read by a machine instead of a designer

A model has to read what you do, where you do it, and who you do it for in plain text, without inferring anything. Established brands fail this test constantly, because years of brand polish strip the plain sentences out.

Fix these 4 things first:

1. State the business, the offer, and the market in plain sentences on your homepage. "We manufacture custom fasteners for aerospace suppliers across the US Midwest" beats "Precision. Partnership. Performance." Plenty of mid-market homepages open with a full-width photo and a tagline, and never once name the category in actual text. A machine reading that page learns nothing it can repeat. 2. Give every service line its own page, titled the way a buyer says it. If customers say "windshield replacement" and your site menu says "glass solutions," the model matches the customer's words and your page loses. One page per offer, plain title, plain first paragraph. 3. Answer the questions buyers actually type. Cost ranges, timelines, what the first engagement looks like, what you need from them to start. Question-and-answer text gets pulled into AI answers constantly because it matches how people ask. A pricing page that says "contact us" gives the model nothing to quote, so it quotes the competitor who published a range. 4. Keep your name, locations, and contact details in text, identical on every page and every profile. A multi-market brand carrying three name variants from old acquisitions reads as three weak entities instead of one strong one. Pick one rendering and enforce it everywhere.

None of this is exotic. Most brands simply never wrote the plain sentences down.

Layer 2: earned mentions, where the heavy lifting happens

Given the 84% figure, the highest-return work sits off your site. Focus in this order:

1. The directories and comparison platforms for your category. For a consumer brand with locations, that starts with a complete Google Business Profile for every location, every field filled, current photos and hours. For a B2B brand, the two or three platforms your buyers check before building a shortlist. AI tools and buyers read those pages constantly, and an empty profile reads as absence. 2. Review platforms, plural. One strong platform is worth less than three decent ones, because the model reads across sources and consistency across sources reads as truth. A brand with a deep profile in one place and nothing anywhere else looks narrower than it is. 3. Trade press, industry associations, partner pages, and "best of" lists. AI tools cite a single mention in a trade publication's roundup far more than another page on your own site. Your partnerships lead already holds these relationships; nobody ever asked them to turn relationships into listed mentions. 4. The places your buyers talk to each other. Reddit, industry forums, private communities. Have your actual experts answer questions under their real names. Humans, and increasingly platforms, spot fake recommendations, and for a brand with revenue to protect, the damage outlasts the gain.

Research from 2026 also found YouTube mentions correlate strongly with AI visibility. A dozen plain videos answering the questions your sales team hears weekly does more here than another gated whitepaper.

On reviews specifically, recency does the work. A steady trickle beats one big push: two new reviews a month, forever, beats fifty in a quarter and then silence. Ask at the moment the customer is happiest, with a direct link, and write replies that carry information. "Thanks Maria, glad the emergency replacement worked out on a Sunday" tells a reader and a machine that you handle emergencies on weekends. Forty replies like that add up to a plain-language description of your entire offer, in your customers' own words, on a platform people read constantly. Almost nobody does this on purpose.

Layer 3: run it like a monthly operating review

You review pipeline monthly. Review this the same way. It takes one owner and 20 minutes.

1. Write the 10 questions your buyer would type. "Best [your category] for [your buyer's situation]." "Who does [service] in [market]." Real phrasing, including the messy variants your sales calls surface. 2. Ask each question in ChatGPT, Perplexity, and Google's AI results, in a logged-out window so your own history doesn't skew it. 3. Log three things per question: did your brand appear, who did appear, and which sources the answer cited.

That third column is the valuable one. The cited sources are a literal to-do list. If AI answers keep citing the same industry directory for your category and you have no presence there, that's next month's task: assign it and budget it.

Then track 4 numbers monthly, in the same deck as your other channel metrics:

1. Appearance rate. Of your 10 test questions, how many named you. Going from 0 to 3 out of 10 is a real result.

2. Referral traffic from AI sources. In Google Analytics, check referrals from chatgpt.com, perplexity.ai, claude.ai, and gemini. The numbers start small. Track the direction, not the size. 3. Conversion rate of that traffic against everything else. Given how much longer AI-referred sessions run, it usually converts better, and that comparison tells you what the channel deserves in budget. 4. New citations found. Count the fresh places your name appeared in the test answers. That's the earned-media work showing up.

Six months of this gives you a trend line you can defend in a board meeting, instead of an anecdote about what ChatGPT said once.

Where the money goes wrong, and your first 30 days

Skip anyone selling guaranteed placement in AI answers. Nobody controls what a model says, the same way nobody controlled Google's first page. The work is real and the guarantee is fiction.

Skip mass-produced content written to game AI systems. The models already read that pattern well. Depth on the questions your buyers actually ask beats volume every time, and a brand with real customers and real experts holds an advantage here that a content farm can't copy.

And skip handing this to whoever has spare time. This channel touches how buyers choose you, so fund it like one: an owner, a monthly review, a line in the budget.

Your first 30 days:

1. Week 1: run the 10 test questions and log what gets cited.
2. Week 2: fix the plain-language layer on your own site. Homepage sentences, one page per offer, real answers to real questions.
3. Week 3: complete the profiles and directories that kept getting cited in week 1.
4. Week 4: set the review cadence and the reply standard, then brief whoever owns trade press on turning relationships into mentions.
5. Then re-run the 10 questions and compare.

The brands getting named ahead of you rarely have anything you can't match. They show up in two or three places you don't, and they wrote the plain sentences you didn't. That gap closes with ordinary work, done deliberately, starting this month.

If you want us to check this properly across search, AI answers, and your local listings, that's exactly what our audit covers at NuroSparx. Send us your site and your category and we'll come back inside 24 hours with what's blocking you. The SEO, GEO and AEO breakdown explains how the three fit together. Run the 10 test questions this week either way, because the citation list they produce is your task list for the quarter. If you'd rather walk the results through with someone who does this monthly, book a call. The habit of checking what AI says about you pairs with checking what AI writes for you.