The problem with B2B AI marketing tactics right now is that everyone’s selling certainty in a system that’s changing weekly.
LinkedIn is drowning in posts about “guaranteed” AI strategies: screenshot after screenshot of viral wins, tools that “definitely” get you ranked, and carousels that “always” perform.
But when you actually test them, most fall apart. Not because the tactics are completely wrong, but because they’re presented without context, caveats, or an honest assessment of what changed since they stopped working.
Over the past couple of weeks, we’ve come across a handful of interesting AI and B2B studies and strategies worth sharing.
In this article, we’ll look at AI tactics that actually work, what’s overhyped, and how to pressure-test similar claims when you see them.
Table of contents
Different LLMs pull from different sources
The biggest mistake you can make as a marketer is to optimize for “AI” like it’s a monolithic thing. It’s not.
ChatGPT Search, Perplexity, and Google AI Mode don’t just rank content differently; they pull from fundamentally different source hierarchies. And those hierarchies shift faster than traditional SEO ever did.
Recent data shows the breakdown:
ChatGPT Search now prioritizes Wikipedia as the dominant #1 cited domain. Reddit dropped to #2. NIH, Medium, Forbes, and arxiv.org appear consistently, while Google.com ranks surprisingly low despite its authority.
Perplexity keeps Reddit as the top-cited source. YouTube and LinkedIn both carry significant weight. Wikipedia appears, but isn’t leading.
Google AI Mode puts LinkedIn at #1, showing a strong bias toward professional content. Reddit stays high, YouTube gets heavily weighted, and Google.com ranks notably higher here than in ChatGPT Search.
The big takeaway is that LLMs aren’t relying on a single set of “universal” sources anymore. Understanding which LLM your audience defaults to actually matters now.
Before you optimize for any of these models, figure out which tool your audience is using. Different audiences default to different platforms, and because each LLM pulls from different sources, your distribution strategy should match where your audience consumes content.
What to do about it
If your audience primarily uses ChatGPT, emphasize structured, factual, reference-style content. Think Wikipedia-grade clarity and citations.
If they rely on Perplexity, create content that lives on social and conversational platforms. Real-time discussions, expert commentary on Reddit, YouTube explainers, and LinkedIn thought leadership. Perplexity pulls heavily from these sources.
If they lean toward Google AI Mode, double down on expert-driven posts, professional pages, and authoritative articles from first-party sources. LinkedIn content performs particularly well here.
Here’s the caveat: AI rankings shift. Fast.
What ranks today might not rank next month. Your advantage comes from adapting faster than everyone else, not from finding some permanent optimization formula.
Track where LLM citations come from in your space.
- Run searches through different LLMs for your key topics.
- See which sources they’re pulling from.
- Then adjust your distribution strategy accordingly.
Google’s preferred source feature: Interesting theory, limited reality
There’s a new ask you can make of customers who love your product: invite them to add you as a Google “Preferred Source.”
This is Google’s latest feature, where users can hand-pick the publishers or brands they trust most, and Google will surface more content from those sources in their search results (big shoutout to Casey Hill for sharing this with us).
It keeps your brand top-of-mind, increases your chances of showing up when they search, and helps you edge out competitors.
WIRED published an article explaining the feature and directly asking their audience to mark them as a preferred source.
The tactical implementation is straightforward. You can reach out to customers via email or article and ask them to add you using this link structure: https://www.google.com/preferences/source?q=[root domain]
Or you can use the AOM method: show a pop-up to every website visitor, asking them as soon as they land.
A stronger iteration triggers the pop-up only for highly engaged visitors: people who visit specific pages, scroll far enough, or spend significant time on site. Targeting them at their most engaged moment increases the likelihood they’ll take action.
These are both smart moves (at least in theory).
The downside
We tested the link, and it led straight to an error page.
Apparently, this feature is currently only available in the U.S. and India, with no rollout timeline for other regions. If your audience is outside those two countries, they can’t use it.
More importantly, from what we’ve observed, this feature mainly influences results in the Top Stories and breaking news sections of Google Search. It doesn’t seem to impact evergreen content, blog posts, or long-tail keywords. This means it’s far more relevant for media outlets than for the average B2B marketer.
The verdict
Unless you’re a news publisher with an audience in the U.S. or India, this feature won’t do much for you right now, but it’s still worth monitoring.
Google may eventually expand it to other regions and content types beyond news. If it ever hits main search results, that’s when things change significantly. But for now, skip it unless you fit that narrow criteria.
The TikTok carousel hype: A case study in validation
LinkedIn is saturated with viral-growth hacks, especially around AI tools. One recent post claimed that using AI to create psychology-themed carousels on TikTok guaranteed millions of views. The author included screenshots of their “most viral” videos as proof.
Normally, we’d scroll past, but the screenshots looked legitimate. So we dug deeper. Here’s how you can validate similar claims yourself.
Step 1: Examine the account data
The screenshots were real. Filtering videos by top views showed plenty of posts crossing the 1 million mark, suggesting that psychology content and carousel format do resonate on TikTok. The setup was simple: basic visuals and facts you could pull from ChatGPT in seconds.
But here’s what the author didn’t mention: most of those viral videos were from last year. When we filtered from “most popular” to “most recent,” a completely different story emerged. Videos posted in 2025 were barely hitting 4,000 views.
The author showcased outliers and last year’s hits, while the AI tactic they’re currently preaching doesn’t seem to be delivering anymore.
Step 2: Run your own AI content validation test
We used Nano Banana to generate AI visuals and focused on Black Friday psychology, breaking down buying behaviors behind it. The whole thing took less than 30 minutes to produce, but it only ended up getting 275 views.
One video isn’t enough to declare the experiment a success or failure. But it brings us back to that LinkedIn post. The author implied that all you need are visuals and psychology facts to rack up views, completely ignoring all the other factors that influence virality.
If you’re expecting to go viral by copying those exact steps: it’s a bust.
Step 3: Understand what works
Spotting hype is good, but figuring out how to turn the idea into something that works for you is better. Many AI tactics online are framed poorly, but still contain valuable nuggets if you’re willing to dig.
With TikTok carousels, factoring in elements that genuinely influence performance can help your content gain traction:
Algorithm changes matter more than tactics. Remember, what worked last year might not work today. So, stay current with platform trends, engage in your niche, watch what’s getting traction, replicate patterns, and keep iterating instead of clinging to tactics that stopped working.
Track slide progression. With carousels, the critical metric is whether people are bailing at slide 1/6 or swiping through to 6/6. More swipes equal better reach (This is your real performance indicator).
Your first slide matters most. Users will scroll past your post in under two seconds. Make sure the visual fits the topic, and the message is bold, clear, and scroll-stopping. If there’s one slide to perfect, it’s that one.
Storytelling drives progression. The best way to keep people swiping is to make each slide intriguing and naturally lead into the next. Don’t give everything away upfront. Deliver the story one slide at a time. We’ve seen carousels with literally zero visuals blow up purely because the narrative pulled people through.
Sound still influences discovery. Trending sounds can give your post an extra push. Audio that matches your content’s mood helps it feel native to the platform.
Validate before you implement
The real skill isn’t finding AI tactics. It’s knowing how to pressure-test them before you waste time.
When you see a viral LinkedIn post claiming guaranteed results, follow this framework:
- Check the account history: Are the “viral” examples recent or from last year? Filter by date, not just by popularity, and remember: outliers aren’t patterns.
- Run a small test yourself: Most AI tools make validation cheap and fast. Spend 30 minutes replicating the tactic and see what actually happens.
- Identify what’s missing from the claim: Most hype posts ignore context: algorithm changes, audience fit, platform-specific behaviors, and content quality beyond the template. Figure out what they’re not telling you.
- Extract the actual insight: Even overhyped tactics usually contain something useful. Separate the framework from the false promises: determine what you can actually use.
This approach works beyond TikTok carousels. Apply it to any AI tactic you see floating around LinkedIn. The ones claiming guaranteed results are usually the ones that need the most scrutiny.
Next steps
- Audit which LLMs your audience uses. Send a simple survey or check analytics for referral sources from ChatGPT, Perplexity, or Google AI Mode. Find out where they’re actually discovering content.
- Map your content to citation sources. Run key searches through different LLMs. Note which domains they’re pulling from in your space. Adjust your distribution strategy to match those platforms.
- Build an AI content validation checklist. Next time you see a viral tactic on LinkedIn, bookmark it. Then check account history, run a small test, and document what happened. Build your own database of what works and what doesn’t in your specific context.
The goal isn’t to dismiss every new AI tactic. It’s to stop implementing blindly and start validating strategically. Most tactics contain something useful; you just have to separate the ones that compound with interest from the hype.
If you’re serious about getting results from AI, join CXL’s live cohorts and learn directly from experts who are shaping what works right now.
→ Build an all-bound marketing engine
→ Competitive Research for Ads Using AI
→ Increase Visibility and Revenue with AI Discovery Engines
→ AI Content Strategy for Changing Search Habits
→ Optimize Pages for AI Search (AEO)
Spots are limited—and so is the window to get ahead.
