Every SEO agency is now claiming they’ll help you rank in large language models. But when you look at their actual plan for doing this, almost none of them are doing anything that will move the needle. Most are just relabeling old SEO tactics as “GEO” or “AIO” and hoping you don’t notice.
Here’s the uncomfortable truth: the agencies selling you AI optimization are usually just parroting advice from LinkedIn posts six months ago. And in a space changing as fast as AI search, that might as well be ancient history.
What Agencies Are Pitching vs. What Actually Works
I recently reviewed about half a dozen proposals from SEO agencies that a prospective client shared with me. Every single one talked about how important it is to show up in ChatGPT and Google’s AI mode. They’re right about that. I think within about a year, traditional search is going to be mostly replaced by AI responses.
But when I looked at what these agencies were actually proposing to do, it became clear they have no idea how to influence the recommendations made by large language models.
Throughout every proposal, the only recommendations were structural changes to website content:
- Adding FAQ sections
- Writing in an “NLP-friendly format”
- Adding structured data
- Creating an llms.txt file
These are fine things to do. But they’re not going to make a big difference.
The Limits of On-Site Optimization for AI Visibility
Let me be clear: basic on-site optimizations are useful but not sufficient. They might give you a 5-10% improvement in AI visibility at most. Research from Onely confirms that these technical tweaks help models parse your content, but they provide only a modest boost on their own.
Here’s why structural changes alone won’t get you recommended: LLMs don’t choose sources based on how you format your page. They choose based on external authority signals. Studies show AI answers are 6.5 times more likely to cite third-party websites than a brand’s own site content.
In other words, who references you matters more than how you structure your page.
If you ask these agencies to show you real data on how their tactics improve visibility in large language models, they won’t be able to provide it. It doesn’t exist.
Why “Best Practices” Don’t Work Anymore
For the last 10 years, SEO agencies could get by with a simple strategy: wait for others to figure out what works, read about it online, then copy it. That was a fine approach when Google’s algorithm changed once or twice a year.
The problem is that things are changing too quickly now. And they don’t show any signs of stopping.
Backlinko’s research found that ChatGPT’s cited sources shifted by approximately 80% in just two months. A tactic that worked last month might be worthless next month. By the time agencies hear about “best practices,” those best practices have already changed.
The GEO tactics I saw in these proposals were clearly just recycled advice from half a year ago. That’s not a strategy. That’s guesswork dressed up in a nice PDF.
What Actually Helps You Rank in Large Language Models
Here’s what an agency should actually be doing if they want to increase your visibility in AI platforms:
Track What Sources LLMs Are Citing
Most importantly, they should be tracking what sources the large language models are looking at before recommending businesses in your industry. The software we use for this is called Peec.ai, but there are now many platforms doing this (Otterly.AI, Profound, and others). They’re all very similar.
Without this data, you’re just guessing where to focus your efforts.
Get Into the Right Directories
Typically, there are only one to three directories that really matter in any given industry. An agency should figure out which ones those are, make sure you’re listed, and then help you climb the rankings so you show up on the automatically generated listicles.
Large language models love these listicles. Research from BrightLocal found that Yelp alone is used as a source in about one-third of AI search results for local businesses. For B2B software, that might be G2 Crowd or Capterra. The specific directories depend on your industry, but the principle is universal.
Identify and Engage With Commonly Cited Discussion Threads
These are typically going to be on Reddit or in Facebook groups. You don’t need to be posting in these forums every day. Typically, the most cited threads are from one or two years ago. You just need to find them and leave some comments that mention your brand in a helpful, authentic way.
This works because LLMs have been trained on forum discussions, and they consider threads with significant engagement to be “trusted sources.” According to Medium’s Reddit LLM Strategy Manual, adding your insight to existing high-value threads lets you piggyback on that thread’s established authority.
The key is authenticity. Five genuine contributions outweigh thousands of spammy ones.
Create Content Across Multiple Platforms
You need to be creating a lot of content, published to the platforms that large language models most often cite. Your content strategy should be informed by the kinds of topics that LLMs tend to cite when people ask about your products or services.
Think YouTube, LinkedIn, Reddit, and industry publications. Analysis of LLM citations found that vendor sites comprised less than 3% of ChatGPT’s citations. It would rather quote a third-party article or forum discussion.
This is why our content repurposing system focuses on distributing your insights across multiple platforms rather than just putting everything on your website.
Consider Press Releases
Press releases can still be effective for AI visibility, particularly for announcements that establish your brand’s credibility and reach.
Why This Is Actually Good News
Here’s the silver lining: most SEO agencies aren’t ready to completely change their entire model. It’s much easier for them to pretend that a few on-site tweaks are all it takes.
That means your competitors are probably using these agencies.
And that’s why there’s so much opportunity right now. Very few agencies are actually doing the work to figure out what works. If you’re serious about getting a first-mover advantage in large language models, you need to be willing to do some experimentation.
The Importance of Experimentation
At least once a month, we run some experiment and get no results from it. That’s just part of the process. But this is how we find the things that actually do work. Then we keep doing more of that and less of the things that don’t.
Any agency that claims to have all the answers for AI optimization is either lying or hasn’t been paying attention. The research is clear: even experts are learning in real-time. What matters is capturing that learning and adapting quickly.
We built our AI SEO service around this principle. Strategy isn’t a document we hand you once and forget about. It’s a continuous process of testing, measuring, and adjusting based on what’s actually working.
How to Evaluate an Agency’s GEO Claims
If you’re evaluating agencies for AI optimization, here are some questions to ask:
What tools do you use to track LLM citations? If they can’t name specific platforms for monitoring which sources AI models cite, they’re guessing.
Can you show me data on how your tactics have improved AI visibility for other clients? Structural changes and llms.txt files are easy to deliver. Measurable improvements in AI recommendations are not.
How often do you test new approaches? If they’re following a static playbook, they’re already behind.
What percentage of your efforts focus on external authority versus on-site optimization? If the answer is mostly on-site, they’re not addressing what actually drives AI recommendations.
The Bottom Line
The shift from traditional search to AI-generated answers is real and accelerating. But most SEO agencies haven’t caught up. They’re selling you the same tactics with new labels.
Getting recommended by large language models requires a fundamentally different approach: tracking what sources AI actually cites, building your presence on those platforms, and continuously experimenting to stay ahead of changes.
If you’re serious about AI visibility, make sure your agency is actually doing the work, not just repackaging old SEO tactics and calling it GEO.
Ready to see what AI-focused marketing actually looks like? We offer a free digital marketing audit that includes an analysis of your current AI visibility and specific recommendations for improvement. No obligation, no credit card required.