Buyers are asking AI about you. Engineers are asking AI about you. AI is asking AI about you.

For an increasing number of prospects, the B2B buying journey now starts in a chat window, and their nascent impression of you is formed before they even get to your website. A buyer who’s never heard of you is building their shortlist inside an AI chat you don’t control and it’s giving answers you can’t see.

And Googling for answers is a rabbit hole you aren’t prepared for. Everyone is trying to spin the data to sell you something, or improve their own AI visibility. The truth lies somewhere between a listicle and an AI hallucination.

I kept digging until I hit bedrock: research data from primary credible sources. This is the real, actual, 100% data-driven State of AI Visibility as of June 2026, volume one. With director’s commentary.

Buckle up.

State I: AI Adoption is Here

The Buyer’s Journey Starts in a Chat Window

The way we research decisions has changed dramatically. From personal to professional questions, instead of asking our network or hitting Google, we open an AI prompt and start asking questions. And more often than not, we take ChatGPT or Claude at their word. They’re just so darn confident and charismatic sometimes. But what are the numbers on this shift? And is AI really informing buying decisions to that degree?

94% of buyers are using AI to research vendors, and 51% of them start that research in an AI chat.

Whether due to preference or corporate mandate, a buyer’s first stop is no longer your inbox, nor your website, it’s ChatGPT. They may not be solely relying on AI for sourcing, but for more than half of buyers, it’s their first stop. To me this reads as: AI visibility aside, trade shows and personal outreach are more important than ever. AI is either warming your leads for you, or taking you out of the running before you even get a chance to hand them a business card.

Data sources: G2 Answer Economy AI Insight Report | Forrester Buyers’ Journey Survey 2025

69% of Buyers Changed Their Mind After Talking to AI.

But wait, there’s more: 1/3rd bought from a vendor they’d not previously heard of, and 85% think more highly of a vendor an AI names. Sit with that for a moment. Somewhere between Air Canada’s chatbot “negligent misrepresentation” and the Chicago Sun-Time’s AI Hallucinated Summer Reading Guide, trust in AI bounced back in a big way.

That’s a little scary right?

What can you do to keep from becoming the next $1 Chevy?

Data source: G2 Answer Economy AI Insight Report

Engineers Use AI More, and Trust It Less

Users with a technical background are on-board with AI adoption. They know that staying up with technology is part of the job. However, as seasoned veterans, they are equally aware that new tech has to earn their trust before it can become truly integrated into their work. The more technical the work, the more opportunities AI has to hallucinate, conflate, or otherwise give you an answer that looks correct, but is, in fact, not.

66% of Engineers visited ChatGPT more than once a month last quarter

With a usage rate already slightly higher than the average American web user, the adoption rate of AI in the technical audience is only going to grow as pressures mount to achieve more at work.

Data source: TREW/GlobalSpec State of Marketing to Engineers 2026, per SparkToro data.

Engineers rate AI answers just 4.7 out of 10 for trustworthiness.

And here is the skepticism we know and love. AI can be an amazingly useful tool, but it’s also a tool that’s already famous for hallucinating answers and telling people to put Elmer’s glue on their pizza. Only 6% of engineers find an AI summary to be sufficient on its own. 94% want more.

Engineers do load-bearing work: in science they trust, and AI isn’t quite there yet.

Data source: TREW/GlobalSpec State of Marketing to Engineers 2026

Don’t panic: 95% of Americans still use traditional search, even as AI usage nearly quintupled from 8% to 38% over the past 2.5 years.

Usage is not exclusive. While AI-tool use has gone up significantly, it hasn’t replaced traditional search, only added to it. The ofttimes tedious workflow looks like this: ask AI a question, then Google each piece of information it comes back with to manually verify.

Data source: SparkToro/Datos

What you need to do

While priority #1 is making sure your website is written for your human visitors, you now have a hard second priority: think of it as a training corpus for AI models. Make sure your site content clearly communicates who you are, what you do, and what you don’t do.

Talk about your people, your expertise, your capabilities, have content attributed to actual SMEs at your organization, and when possible attach hard numbers to any claim.

Bottom line: be specific.

Giving AI factual grounding is your first line of defense against becoming another confident hallucination.

State II: Every AI, Every Model, Every Response, is Different

Generative AI Services Popularity according to Cloudflare

To provide the most honest state of the distribution of traffic across AI platforms, we pulled current data (as of June 29th, 2026) from Cloudflare Radar Generative AI Services Popularity.

1. ChatGPT
2. Claude
3. Gemini
4. Perplexity
5. Deepseek
6. Grok

If you filter by industry you get some variation of this list: ChatGPT on top, and the order of the others varies by industry and audience.

With ChatGPT being the first AI most of us used, and the fact that it powers Microsoft Copilot, it’s not surprising to see it take the #1 spot. Claude has also risen in popularity, particularly with technical audiences due to increased usage of Claude Code.

Worth noting regarding Gemini (Google) coming in #3: that does not include their ubiquitous AI summaries in search results, the inclusion of which would no doubt increase their rankings.

Every response is a roll of the dice.

Ask three AIs for a recommendation, and you’ll get three (sometimes wildly) different answers. Each AI prioritizes different sources, has different crawling behavior, and hallucinates with a different level of confidence. It’s important to understand exactly how much variance you can get, not only between different models, but within the same model even asking the same question.

Recommendation overlap between AIs is almost non-existent.

Across ChatGPT, Perplexity, Claude, Gemini, and Google AI Mode, seven in ten sources were cited by only one AI. Meaning between five AIs all asked the same question, only three sites showed up in multiple answers. The other seven were unique to each AI’s response.

Talk about casting a wide net.

In the competition for AI visibility, you’re vying for attention from platforms with divergent standards for citation. Perplexity in particular has a penchant for sourcing citations from Reddit and third-party content, meanwhile ChatGPT often uses Bing for search grounding. On the flip side, Claude almost never cites YouTube or Reddit, preferring documentation, vendor pages, and editorial sites.

According to research on 127,198 AI citations across five engines done by SurfacedBy, the amount of citation varies by model too: Gemini listed 11.0 sources per answer on average. ChatGPT listed 3.7. Perplexity listed 8.6, Google AI Mode (7.8), and Claude (6.8) sat in between.

Data Source: SurfacedBy – We Analyzed 127,198 AI Citations. The Five Engines Barely Read the Same Web.

<1 in 100 Chance that AI will give the same top list twice in 100 runs.

AI’s “thinking” is essentially a very advanced predictive text generator, meaning it’s based on the probability of what might come next. Every query is a fresh prediction, a new roll of the dice. You’re fighting for visibility in a system where the arbitrator doesn’t even agree with itself.

Pair this with the previous finding in the level of variance between models and suddenly AI visibility starts to feel like trying to hit a bullet in mid-flight with another bullet.

The Matrix has you.

Data source: TREW/GlobalSpec State of Marketing to Engineers 2026, per SparkToro data.

What you need to do

If reading this has not only confused you, but also given you a case of tachycardia, you’re not alone. There are support groups on LinkedIn.

Let’s break it down, and keep it (relatively) simple:

• ChatGPT is the clear front runner, and since it powers Copilot, it’s bundled with Windows. If you have to pick one AI to focus on, focus on ChatGPT.
• Get AI analytics for your site. These can be hard to come by because AI crawlers skip some common tracking methods, but if you use Cloudflare you’re ahead of the game.
• Ask your sales and account reps about questions they get all the time. Start feeding those into ChatGPT and see if it’s able to answer them. Ask it what it knows about your company, what you specialize in, what you don’t do, who some key staff are, what your capabilities are, and so on. If it can’t answer direct questions accurately, it’s time to start wrenching on your site content.

Next: Signal vs. Noise and What Actually Works

Now that you have a basic grasp of the landscape of AI visibility, we can dig even deeper, to those burning questions that keep marketers and the C-suite alike up at night:

What signals are AI looking for?

When they visit your site, what are they actually seeing?

And most urgently: with everyone and their cousin trying to make a buck off of AI-induced tachycardia, how do we know what’s hype, and what actually works?