Your rankings are holding. Your clicks aren’t.
Google’s AI options reply the query on the outcomes web page, and the searcher by no means leaves. From January to April 2026, 68% of U.S. Google searches ended and not using a click on. And fewer of Google’s site visitors reaches the open internet than it used to.
As a result of for a rising share of B2B consumers, the search doesn’t begin on Google in any respect. It begins in an AI chatbot.
Google’s outcomes web page, its AI options, and the standalone LLMs all pull from a mixture of sources. Reddit threads. YouTube movies. Assessment profiles. LinkedIn posts. Comparability pages.
That’s your natural development stack. It’s additionally the combo your consumers use to construct a shortlist.
If you happen to solely optimize your website, you’re competing for a shrinking slice of the pool. To remain aggressive, website positioning has to increase past on-site.
At the moment, I’ll present you which ones sources every floor leans on and what to construct first.
Search Is Splintering Throughout Surfaces
For our State of B2B AI Discovery 2026 report with AirOps, we tracked roughly 380,000 AI solutions overlaying 81 enterprise software program subjects throughout six AI surfaces. Until linked in any other case, the figures beneath come from this research.
Let’s begin with what didn’t change. Product pages nonetheless get cited greater than anything.
Throughout each AI floor we tracked, vendor websites, integrators, and power directories took the most important share of citations. That ranged from 52% on Perplexity to 72% on Gemini. Assessment websites added one other 9% to 21%.
Keep in mind: Vendor share isn’t yours by default. It contains your opponents’ websites and the software program directories that listing you side-by-side.
That is the stack’s basis, and it’s crowded. Your product pages compete together with your rivals’ pages. The comparability pages written about you compete with those you write your self.
The layers above the inspiration are the place search begins to splinter.
Google’s AI Overviews and AI Mode floor their solutions in reside Google Search outcomes. They inherit what already ranks. ChatGPT, Claude, Gemini, and Perplexity run their very own retrieval.
The citations cut up proper alongside that line.
Google’s two AI surfaces put 7% of their citations into social platforms like YouTube and LinkedIn. ChatGPT, Gemini, and Claude sit at 0.3% or beneath.
Area by area, the hole will get sharper:
- Reddit seems in 52% of AI Mode’s citing solutions and 39% of AI Overviews. ChatGPT cites it in 21%, and Claude didn’t cite it in any respect.
- YouTube seems in 47% of AI Mode solutions and 38% of AI Overviews. ChatGPT and Claude: basically zero.
- LinkedIn reaches 13% of AI Mode solutions, 9% of AI Overviews, and 6% of Perplexity’s.
- G2 lands between 23% and 30% on 5 of the six surfaces. It’s the one area each floor agrees on.
So “AI visibility” is at the least two video games.
A model may be in every single place in Google’s AI solutions however invisible in ChatGPT and the opposite standalone fashions. Or the reverse. Utilizing a blended visibility rating hides each’s contribution.
See additionally: Wikipedia, Podcasts & Reddit Now Drive AI Citations, Ask Your PR Workforce How They Did It
Construct Your Natural Progress Stack Layer By Layer
Natural visibility now comes all the way down to floor protection: exhibiting up on the domains every AI floor trusts most.
Your new natural development stack has to handle each internets. Right here’s the right way to do it.
See additionally: What Search Engines Belief Now: Authority, Freshness & First-Celebration Indicators
1. Begin With The Pages You Personal
AI instruments are catching on quick with B2B consumers. In keeping with G2, 51% of B2B software program consumers now begin their analysis in an AI chatbot extra typically than Google, up from 29% a 12 months earlier.
However Google nonetheless dominates search general. It handles roughly three-quarters of all desktop searches. AI Overviews alone attain 2.5 billion month-to-month customers and one other billion use AI Mode.
I nonetheless suppose rating issues. It simply doesn’t assure a quotation anymore. AI Overviews have been pulling fewer citations from top-ranking pages.
The standalone engines lean on product pages too.
Gemini sends 72% of its citations to vendor and product domains, whereas G2 exhibits up in simply 14% of its solutions. Gemini is selecting between vendor pages, and yours are up in opposition to everybody else’s in your class.
Then there’s comparability content material.
Eighty-four of the 100 most-cited URLs in our knowledge have been both comparability pages (59) or listicles (25). Collectively, they drew over 66,000 citations. Whoever writes these pages shapes how AI frames your class.
Loads of groups pour funds into the weblog and let product and comparability pages go stale. I’d flip that. Repair the pages AI already cites earlier than you write one other put up.
As James Scherer, Basis’s VP of Technique, places it: “Editorial will get the funds; product pages get the citations.”
Motion: Determine the ten most-cited comparability pages and listicles in your class. Test the way you’re framed in each. Then publish your personal comparability web page in opposition to your prime three opponents.
2. Earn A Place In The Communities Your Consumers Frequent
Reddit was the most-cited exterior area in our dataset, incomes 159,000 citations throughout 12,500 distinctive URLs.
Reddit’s logged-out customers, who largely arrive from serps, grew 27% 12 months over 12 months in Q2 2026. They now outnumber logged-in customers, 77.7 million to 52.6 million. Threads rank on Google. Google sends readers into the threads. Then Google’s AI options cite them.
The layers feed one another.
However these layers also can shift quick. After reviews that ChatGPT was citing Reddit much less, we prolonged our monitoring and located its Reddit quotation fee fell from 21% in July to 10% in August, whereas Google’s surfaces stored citing Reddit.
How a lot group platforms like Reddit matter additionally is dependent upon your class. Neighborhood citations ranged from 5% of the full for website positioning platforms to 14% for EDR/XDR safety instruments.
If you happen to’re new to the platform, begin with this SEJ AMA with a Reddit advertising veteran.
My largest recommendation: Don’t deal with Reddit like a launch channel. I’d commerce one viral thread for regular, helpful solutions throughout 50.
Motion: Unfold your effort throughout many threads and subreddits. Don’t wager on one. Monitor group citations weekly, as a result of month-to-month reporting would have missed a drop like ChatGPT’s.
3. Present Up The place Classes Are Outlined
LinkedIn is a Google-surface play. It reaches 13% of AI Mode solutions, 9% of AI Overviews, and 6% of Perplexity’s. ChatGPT, Gemini, and Claude didn’t cite it in any respect.
Throughout B2B software program classes, social is the second-largest quotation supply after model and product content material, at 6%.
LinkedIn is the place executives and practitioners set the class narrative. Google’s AI options choose up that narrative. It’s why worker amplification belongs within the stack.
Don’t fall into the likes and follows entice. If the objective is getting cited, watch whether or not Google’s AI solutions choose up the best way your folks describe the class.
Motion: Decide the class questions you need to personal. Publish on them each week from executives and practitioners. Measure the affect in AI Overviews and AI Mode, since that’s the place LinkedIn will get cited.
4. Make Video Content material For Google’s AI Search Options
YouTube exhibits up in 47% of AI Mode’s citing solutions and 38% of AI Overviews.
It’s tempting to learn that as Google favoring its personal property, however our knowledge complicates that. Gemini can be a Google product, and but YouTube confirmed up in 0.3% of its citations: roughly 17x lower than on AI Overviews.
The distinction is how every floor finds its sources. AI Overviews and AI Mode are Search options. Google makes use of question fan-out, operating a number of associated searches to construct every reply, so movies that already rank in Search get pulled in.
Loads of B2B movies are made for individuals who already know your model and product. However the movies that earn a spot on this layer reply the questions consumers ask earlier than they’ve picked a vendor.
Motion: Ship comparability, alternate options, and best-of movies. Title them the best way your consumers phrase the query.
5. Deal with Assessment Profiles As Retrieval Infrastructure
The standalone engines lean arduous on evaluate websites. Claude and Perplexity draw about one in 5 citations from evaluate domains.
Claude’s third-party citations are nearly fully evaluate platforms. G2 seems in 30% of the Claude solutions that cite sources. Capterra seems in 13%, and Gartner and PeerSpot in about 9% every.
Class issues right here too. Assessment citations ranged from 8% in managed databases to 25% in ATS and recruiting software program.
Consumers again this up. In G2’s 2026 analysis, evaluate websites (38%) narrowly handed AI chatbots (37%) as the highest supply shaping B2B software program shortlists.
G2 isn’t a one-time push. For Claude and Perplexity, your profile is a reside supply, so run it like one.
Motion: Give your G2 profile a named proprietor and a month-to-month evaluate goal.
What To Construct First
No one builds each layer directly. Run two checks earlier than you begin.
- Ask consumers which AI instruments they use. Do it on gross sales calls and through onboarding. In the event that they use Google, prioritize the group, video, and social layers. In the event that they use ChatGPT or Claude, prioritize your personal pages and evaluate profiles.
- Pull your class’s supply combine earlier than you set a funds. Assessment citations swing 3.3x throughout classes. Neighborhood citations swing 2.5x.
Copy the improper natural development stack, and you find yourself investing within the improper channels. The AirOps knowledge exhibits simply how a lot the supply combine adjustments from business to business, even inside the SaaS house.
The managed database area of interest is an illustrative instance. A model on this house that borrows an HR tech playbook pours funds into evaluate websites, which earn simply 8% of that class’s citations. In the meantime, it starves communities, which earn 13%.
As soon as you understand which layers matter in your class, match them to the group you’ve. At each measurement, the query is similar: which web are you continue to lacking?
- If it’s simply you: Your personal pages, plus the one layer your class leans on most. The bottom layer will get cited on each AI floor, so it pays off whichever software your purchaser makes use of. A second layer you possibly can’t sustain with simply goes stale.
- Three to 5 folks: Add a second layer from the opposite web. If you happen to began with Reddit, which feeds Google’s AI solutions, add evaluate profiles, which feed the standalone engines. A second layer on the identical facet solely makes you stronger the place you’re already seen. Give evaluate profiles a named proprietor, as a result of they go stale quick.
- Eight or extra: One proprietor per layer, plus one one who reviews citations by floor. At this measurement, the larger threat is drift. A blended rating hides it when one platform adjustments what it cites, the best way ChatGPT did with Reddit.
Lastly, measure by floor.
Run a hard and fast set of 100+ prompts on a schedule. Monitor quotation fee and share for every AI floor. Set an alert for any 20% week-over-week swing.
Watch sentiment per platform, too. On the identical prompts, AirOps’ 100-point sentiment scores ranged from 81 on Perplexity to 69 on Claude.
Then tie it again to the numbers management already watches, like branded searches, referral site visitors, and pipeline sourced out of your content material.
Give it 12 months.
The primary quarter is sluggish. You’re fixing outdated comparability pages and setting baselines no person requested for.
By month 12, it compounds. Your comparability pages are those AI quotes in your class. Your evaluate profiles keep recent and not using a scramble. Your threads and movies present up when Google solutions your consumers’ questions. And when a platform adjustments what it cites, you catch it that week.
Final month, you constructed the engine. Now you understand the place to level it.
Search will get you discovered. The stack will get you chosen.
Extra Sources:
Featured Picture: Roman Samborskyi/Shutterstock
