From adoption to influence.
Welcome back. In part 1 of The (Real) State of AI Visibility we talked about how professionals in the manufacturing sector are adopting AI, and the vast variance between models in how they respond to users’ questions. TL;DR: adoption is up, trust is middling, and each model is its own flavor of black box.
With all of that laid out, we can start looking at how AI perceives you, and what you can do to influence that perception. So let’s dig into the meat of the conversation: what does AI see as signal vs. noise, and what actually works to improve your visibility.
I’m taking the same approach as last time: looking at primary sources for original research, no wishful-thinking or marketing fluff allowed. Just the facts please.
State III: When AI is looking at you, what does it see?
AI visibility is about more than being cited in AI recommendations—though that is the ultimate goal. It’s about what AI can see about you when it looks for you on the web. Can it see you at all? And if it can: does it like what it sees? Is your content written and architected in a way that is easy for LLMs to understand?
Can AI crawlers see your website?
According to the research the #1 citation signal is as dead simple as it gets: before schema, content, or off-site strategy can work, AI crawlers need one thing: to be able to reach the pages on your site and read the content. I know, “duhhhh, of course”. But with new AI crawlers rolling out all the time and companies like Cloudflare rolling out new AI-scraper protection, there are plenty of ways for AI traffic to get blocked without you even realizing it.
When AI first started gaining momentum and the copyright issues were top of mind, many security plugins and platforms implemented protections to prevent AI from training on your content, some which may have been activated automatically. The lesson here is don’t assume that your site is accessible to crawlers, or you might be in for a rude awakening.
What to do: Run a crawlability checker on your site to see if you’re blocking anything that you didn’t intend to.
Author’s note: as of June 2026, users on Cloudflare’s free tier with bot protection enabled may be blocking ALL bot traffic, including AI crawlers and fetchers.
Data source: AI Citation Ranking Factors Analysis | Zyppy
Javascript rendered content is invisible to every AI except Gemini.
Here’s another way that your site might be invisible to AI crawlers without you even realizing it. Google’s search crawler is the only bot we know of that is going to render javascript on the page. So if your site renders content client-side, meaning parts of it load dynamically, or have a loading spinner when the site is first loading, it may be invisible to AI crawlers. This isn’t an issue for most folks, but there was a time when some page builders and enterprise content management systems did lean on this. So if your site was built between 2016 and 2020, this is something to look into.
What to do: Turn off Javascript in your browser and load your website: this is what AI crawlers see.
Data source: The Rise of the AI Crawler | Vercel + MERJ
Can AI crawlers read your website?
Okay, your site security let the bot through, your content is all loading without javascript—what does the crawler actually see? Well, it varies. ChatGPT, for example, sometimes grabs page content in chunks and makes inferences based on that to generate its answers. This means that just because content is on your page, and ChatGPT crawls that page, it may not actually read the part you want it to. Other LLMs on the other hand, may read the whole page when it generates an answer, using more tokens, but results in more accurate and informed results for complex topics.
Based on Dan Petrovic’s testing against OpenAI’s Assistants API web search tool an AI crawler generating answers for a user follows these steps:
- Fetch the URL
- Strip out anything that isn’t page content: HTML, Javascript, Structured Schema, etc.
- Generate a Markdown (MD) version of the page
- Read some amount of the content (the entirety or a small sample)
- Generate the answer in the user’s chat window
This is a big one, because the end result is a simplified version of your content being consumed, which may or may not capture the information you consider most important for the AI to know.
Data source: How ChatGPT Sees the Web | Dan Petrovic
Heading structure and semantic markup are a must.
So if crawlers are reading content with an uncomfortable level of inconsistency, how can you improve AI crawler consumption? Research is showing that having a correct heading hierarchy (h1, h2, h3, h4) is key. AI relies on content being well structured, with descriptive headlines, to optimize its ability to understand what it’s consuming. On blog posts this is easily accomplished, however page builders can sometimes lead to inconsistent implementation of semantic conventions, which can get you into trouble with this one.
What this means: Run key pages on your site through a heading checker tool and look at the results: if you can get the gist of the content from those headings in the order they’re presented, you’re in good shape. If not- you have work to do.
Data source: How to Optimize for ChatGPT | SERanking
Markdown is for efficiency, not visibility.
Earlier I mentioned that AI crawlers convert your content into Markdown for consumption as part of their process. One thing you can do to help improve efficiency is offer MD versions of your site content that crawlers can request directly, thereby speeding up their crawl. Additionally, Cloudflare’s Markdown Negotiation feature also embeds structured schema data in the MD it generates, which can be a great boon for you if you’ve invested the time in implementing robust structured schema on your site.
What this means: A nice to have, not a need to have. So far.
Data sources: Agentic Engine Optimization | Addy Osmani
Most AI crawlers do not read llms.txt.
You may have noticed a new setting to enable llms.txt in your SEO plugin in the last 6 months, this is a “community standard”, ie something people are trying out but it doesn’t have any official sanction or adoption by AI platforms. Essentially llms.txt is a summary of your site content, pulled from the database, and smushed into one long text file. Is it pretty? Not really. Does it give AI an accurate snapshot of your site content in one easily readable file? Also, not usually. Not unless you write it yourself, because automatically generated versions often contain an abundance of irrelevant content like Thank You pages, Terms and Conditions, etc.
The data isn’t looking good: according to SE Ranking across ~300,000 domains there’s no correlation between AI citation and llms.txt. My own observation of llms.txt consumption on our client sites tells the same story: sometimes training crawlers grab it, but the rate of consumption is currently miniscule.
What this means: llms.txt isn’t the standard we thought it might be, but the story is ongoing.
Data source: How to Optimize for ChatGPT | SERanking
Audit your site to make sure it’s accessible and legible for AI crawlers.
Let’s put some items on your todo list to make sure that you’re hitting the fundamentals for AI visibility with an audit of your website. The effort is minimal, but depending on the results the payoff could be real.
- Start with a crawlability check to see if your site is blocking crawlers in a way that is impeding AI crawlers’ ability to access your site.
- Then check Cloudflare’s Agentic Readiness Scan, and only worry about the first part of results for now: can the scan access your site, is your robots.txt good, and is your sitemap accessible. If you have markdown negotiation, then you are probably using Cloudflare for DNS, so good on you.
- Turn off javascript in your browser and navigate your site to see what content might be invisible to AI crawlers without you realizing it.
- Run your key pages (homepage, services, products, blog) through a heading checker tool to confirm that your headings are present, in the correct hierarchy (H1 -> H2 -> H3, etc), and descriptive.
State IV: AI is looking for an outside opinion
The biggest thing you can do to improve AI visibility is not on your website. LLMs want to see that you have a presence outside of your website, and that you’re mentioned favorably. Different AI models weigh external presence differently, and even weigh the sources differently (for example Perplexity really loves Reddit).
Just like building your backlink portfolio in the golden era of SEO, your goal should be to distribute content outside of your own website, and get mentioned positively by publishers relevant to your industry. But before you do that, the lowest hanging fruit is to make sure to get your site listed in industry directories, or update the profiles you already have so they align with what you say about yourself on your website.
Ranking does not mean recommended.
Fan-out searches are when AI runs one search, and then based on those results, performs additional searches to dig deeper, resulting in a richer and more varied result set than a single search can provide. So much so that research by Zhang, et al, found that from 55,936 queries across six LLM search engines and two traditional search engines, 37% of the domains cited were unique to LLM-based results. AI search draws from a measurably more diverse set of domains than traditional search, so while ranking in traditional search is always going to be a valuable asset, it’s not the whole story.
What this means: AI is searching- so search presence is key, however it’s searching a lot, so even if you’re not on the first page of results, that doesn’t mean you won’t get cited. Content about niche subjects may not get you to page 1 on Google, but it’s gold for getting you cited by AI.
Data source: Does LLMs.txt impact your AI visibility and citations? | SERanking
To AI, Mentions = Credibility
One thing every piece of research on the subject I’ve seen can agree on is that AI is looking for mentions of your brand in the wild to validate your presence and expertise. Backlinks are great, but what they really want to see is people speaking favorably of you, with or without the corresponding link to your site. In fact, in Ahref’s study of AI visibility factors, mentions outperformed backlinks by ~3x! This is an area where grey-hat practices are seeing a resurgence, similar to SEO link farms circa 2013. At the rate AI tools are evolving, I wouldn’t expect these tactics to remain viable in 2027, so be wary of anyone trying to sell you listicle placements for AI visibility.
What to do: Using Google Search Console, Bing Search Console, or third party tools like SEMRush, review your backlink portfolio. The more uniquely linking domains the better, having a multitude of links that all come from the same domain quickly depreciate in value.
Data source: An Analysis of AI Overview Brand Visibility Factors | Ahrefs
Presence on third party directory/review sites (ThomasNet, etc) resulted in up to 3x more citations.
AI loves authority. It may not always know what signals true authority, but they love it anyway. Third party directory listing and review sites are the lowest hanging fruit for you to start improving your AI citations today. Especially if those sites allow you to collect reviews and ratings, AI will eat that up.
What this means: Get your business into industry directories, and/or update your information in the directories you are in. For bonus points add them to your site’s structured schema to strengthen your entity resolution— AI strips out structured schema when they consume content, but search crawlers still read it to understand your digital footprint.
Data source: How to Optimize for ChatGPT | SERanking
Brands with a smaller digital footprint benefit the most from distributing articles through third-party outlets.
When the same piece of content was distributed through third-party publishers instead of self-published, brands saw a median 239% increase in AI citations — and the effect was strongest for brands with the smallest existing digital footprint. Translation: if you are a mid-size manufacturing company, leverage your relationships, memberships, and trade publications to amplify your message. You need robust content on your site that demonstrates expertise, but you also need to get that expertise out in the world to build the authority you need to grow both your digital footprint, and your visibility to AI.
What this means: Press releases, guest articles on membership orgs and trade publications- if it’s industry relevant and you can get the story/article posted there, do it.Data source: Coverage Breadth: The Latest GEO Research on Brand Visibility | Stacker‘
State V: AI-driven referred traffic is low volume, high intent.
We call AI referred traffic, when a user clicks a link to your site from their AI chat, “the high intent trickle”. It won’t be a huge proportion of your overall site traffic, but because it’s been pre-qualified by the AI chat conversation, it can convert at a much higher rate than organic or branded traffic. This ties back into Part 1 where we talked about procurement oriented users looking favorably on companies mentioned by AI, and even feeling comfortable trying a new vendor based solely on AI’s recommendation.
Analysis by Orbit Media Interactive of 97 sites across 29M sessions through June 2026 found that visitors referred by AI convert far above organic: roughly 1.91%, which was nearly 3x organic traffic sources. Drilling into specific models, ChatGPT converted at 2.1% vs Google at 0.5%. Meanwhile Ahrefs found that AI driven visitors were 0.5% of traffic, but 12.1% of conversions.
This shift in how users make decisions is changing how we look at traffic analytics. As users get directly referred by AI, or land on your site from a branded search after AI mentions you, the value of different types of traffic is shifting. Tracking what traffic actually converts and what influenced that conversion has always been both the most important, and most difficult to accomplish, task for digital marketers when it comes to proving our value.
These findings align with the earlier cited survey results about the rise of AI in the buyer journey and paint an interesting picture: your traffic may steadily decline, but the traffic you do get will be the highest-intent, bottom-of-funnel buyers.
Data source:
AI Traffic Conversion Rates: New Research from 97 B2B Websites and 29M visits | Orbit Media StudiosAn Analysis of AI Overview Brand Visibility Factors | Ahrefs
Next: What auditing 500 manufacturing company websites showed us about AI visibility.
In Part 1 we talked about adoption, skepticism, and the extreme variability in AI responses to user queries. In Part 2 we covered some common pitfalls that could be hampering your visibility to AI crawlers, what they look for to determine authority and cite-ability, and the way the “high intent trickle” is changing website traffic patterns.
Are you ready to go one step further? We were too, so we designed our own research study focused exclusively on the manufacturing industry.
We curated a list of 500 manufacturing companies across the US, focusing on the mid-market segment. We organized them by niche— each with specialized prompts, to better emulate actual user queries. We audited their site infrastructure, content freshness, structured schema, visible certifications, and more, to assign them a composite readiness score. And then the moment of truth: we ran thousands of queries using our specialized prompts to drill into what factors influence AI visibility and citation for manufacturing companies.
Will we peel back the curtain and uncover hidden truths about AI visibility, or will we be left with more questions than answers? Stay tuned for Part 3 and find out what the research showed us.
Who are we? That’s a question we can answer from the head or the heart. Here’s the brainy answer: We are a digital agency offering holistic solutions that integrate custom web functionality with multichannel marketing strategies. We take an idea from a sketch on paper all the way through getting it online and in front of people. We help our clients create comprehensive online marketing plans from inception to completion. But the heart of the matter is this: We founded build/create to be a place where people—employees and clients—can thrive. That means protecting the sanctity of work/life balance, giving our team the space to do their best work, and thinking beyond our clients’ needs to the needs of their customers. We’re here for the long term, and our values reflect that commitment.
