We published the original version of this case study in November 2025. A lot has happened since then. It is time for an updated account of what we’ve done, what has compounded, and what the results look like now.
1. Context: Practicing What We Preach
By mid-2025, it became clear to the team at Proper Propaganda that AI search engines like ChatGPT, Google AI Overviews, and Perplexity had become discovery gateways for buyers of our services. So in July 2025, we turned our GEO framework inward. If we were going to help clients win in AI search, we needed to lead by example and show up ourselves. This updated case study details the full volume of tactics we have tried and where things stand today.
2. Defining the Battlefield
We began with a diagnostic phase back in July 2025 and modeled actions exactly on the framework we built for clients.
At the outset, we identified 100 prompts prospective clients might use to find agencies like ours — ranging from “What is the best PR agency for consumer tech?” to “How can we get products featured in AI search?” Our prompt universe has since grown and we now monitor 695 prompts generating 15,647 responses across five AI platforms (ChatGPT, Perplexity, Gemini, Google AI Mode and AI Overviews).
We benchmarked Proper Propaganda against both niche consumer tech agencies (like Max Borges Agency and Spark) and giants (like Edelman and Walker Sands). We used Scrunch, our preferred AI visibility monitoring platform, to establish our baseline presence across our chosen LLMs and to provide an intelligence level to our program.
Result at the baseline: Zero visibility. We didn’t appear in any AI answers. That was the starting point.
3. Fixing the Foundation
Our first major move was a technical and content overhaul of our site. Working with our web developer, we restructured pages to align with the S-E-T framework (Structure, Explainability, Trustworthiness).
Key updates included:
• Creating an FAQ hub built from real AI prompt phrasing (“How can PR influence AI search?”, “What is share of scrape?” etc.)
• Applying structured markup and clear H1/H2 hierarchies for crawlability
• Updating metadata and internal links to support AI-friendly entity clustering
• Rewriting existing service and blog pages with short paragraphs, literal explanations, and cited examples
We treated the site like a knowledge graph for bots that was also readable by humans. Bots now make up the majority of internet traffic and, thanks to Scrunch’s bot monitoring features, we know that they exceed human visitors to our site by almost 100x. While we still think our site is pretty and has appeal for humans, the main visitors do not smoke, drink, or breathe.
4. Building Anchor Hubs
As our GEO programme matured, we developed a more deliberate content architecture around anchor hub pages. These were based loosely on an outline from Gini Dietrich at Spin Sucks and some other conversations we had with SEOs in our network.
Anchor hubs are deep, structured pages designed to be the citable and authoritative resources on a topic and category. Each anchor hub is built as an evidence library: it has clear definitions, original frameworks when possible, embedded data, third-party citations, and FAQ sections that mirror the natural-language queries AI platforms receive. As an example, take a look at our consumer tech category lifecycle framework.
The anchor hub approach is now central to how we think about our site. The goal is for every major topic we want to own in AI search to have a page that is, unambiguously, the best thing on the internet on that subject in our category. More hubs are under development as I write this.
5. Treating the Website as a Media Outlet for Bots
One of the most important conceptual shifts we made was deciding that a meaningful portion of our website should function as a media outlet for AI bots. This reframe changed how we approach content production.
Borrowing from the famous book by Daniel Kahneman, we call our approach “content fast and content slow.”
Despite all the noise from Google about the importance of writing by humans for humans, the tech giant has de-indexed some of our more popular human-written content; stuff that performed really well on places like LinkedIn and Substack. Indeed, some of our most heavily-scraped content has been written by bots, and is made for bot, not human, consumption. To be clear, we aren’t running a slop farm, and we are not suggesting you should either. Indeed we deploy a healthy mix of both kinds of content. Our experience reinforces the idea that AI systems often scrape structurally clear, machine-legible content more willingly than emotionally compelling editorial prose.
In terms of measurement, we monitor AI agent traffic daily in Scrunch. This tells us which AI crawlers are hitting which pages, how frequently, and what they appear to be indexing. It is the closest thing we have to a real-time feedback loop on what the machines are reading. We use it to make decisions about where to invest content effort and which pages need structural reinforcement.
6. Content Refinement and Third-Party Seeding
Our thinking here was formed by a few interesting data points. We came across great research from Stacker and Scrunch that show a median 239% lift in AI citations when third party sources publish or syndicate content from a brand. The issue was, in our industry, these opportunities are not as plentiful as in others with reams of regular media coverage. It is a common problem for B2B companies.
Because of that, we put a lot of effort into refining the owned channel. Specifically, we:
• Updated and optimised key blog posts using GEO best practices (natural-language Q&A, schema-friendly formatting, quotable statements)
• Launched our Resource Centre page, featuring tools, podcasts, case studies, and stats sheets
• Syndicated content to Substack and my LinkedIn, particularly as long-form articles, which are more aggressively scraped by certain LLMs.
• Where and when we could, we secured citations on trusted external sources, increasing the number of third-party mentions that AI engines reference when responding to GEO-related prompts
7. Podcast Tour
Since last summer, I have been on a bit of a podcast tour. The intersection of PR and GEO has been the primary topic. I’ve been on 16 shows, among them the world’s largest PR pod. The rationale was straightforward: podcast transcripts, show notes, and associated articles are indexed by AI systems. A founder making the same clear, consistent arguments about GEO across multiple high-authority external platforms builds the kind of distributed, corroborated signal that AI engines use to validate expertise and authority.
As many podcasts also run their content on YouTube – which is heavily scraped by the Google LLMs – pod appearances can really help presence on these important platforms.
8. Monitoring and Iteration
With Scrunch as our dashboard, we track share of answer across engines weekly and monthly, and monitor agent traffic daily. We use the same methodology we apply to client programmes: visibility benchmarking, sentiment review, position tracking, and citation source analysis. The daily agent monitoring in particular has become an essential operational habit, as it tells us what the machines are actually reading and gives us a feedback loop that would otherwise take months of guesswork to replicate.
9. Results
The gains from the original case study have continued to compound. The slow start we described in November 2025 — where meaningful traction took 14 weeks to materialise — has given way to a programme that’s delivering large for us. Here is where things stand as of May 2026.
| Metric | Result | Notes |
| Overall AI Share of Answer | #1 at 16% | Ahead of Edelman (13%), Highwire (11%), Hotwire (9%) |
| Platform-Level Share of Answer | #1 on ChatGPT (19%), #1 on Perplexity, #1 on Gemini (tied, 12%) | ChatGPT: leads Edelman by 2 points. Gemini: tied with Edelman, ahead of all others. Perplexity: category leader. 100% positive on Gemini; 99% on ChatGPT. |
| Position Quality | 77% of the time we are in the top position in AI results | 23% middle, 0% bottom. +7% improvement over last 12 weeks |
| Sentiment | 100% positive | Zero mixed or negative mentions across all monitored platforms |
| Citation Growth | +34% overall / +221% on Gemini | Gemini citations up 221% in last 12 weeks. |
| Influence Score vs. Competitors | 31.5 vs. 9.6 (nearest competitor) | properpropaganda.net scores 31.5 across 9,408 tracked domains. Edelman scores 0.95. Walker Sands scores 1.3. |
| Citation Consistency | 0.335 — highest of any domain with meaningful volume | Higher than LinkedIn (0.22), Reddit (0.17), Forbes (0.16). Most reliably cited source in the entire tracked universe. |
| Prompts Monitored | 695 prompts and 15,647 AI responses | Across 5 platforms: ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI mode |
| New Business / Leads | Significant closed revenue. 5x inbound lead volume increase | Multiple clients signed via AI search discovery. Top of funnel leads have grown by 5x |
| Earned Media | 16 podcast appearances | Including the largest PR podcast globally. All centred on AI search. |
10. What the Citation Data Shows
We track citation performance across every source AI systems use when responding to prompts in our category. This is 9,408 domains in total. properpropaganda.net has an Influence Score of 31.5, placing it in a virtual tie with Linkedin.com (31.7) for the top spot across the entire domain universe. Every other agency in the dataset — including Edelman, Walker Sands, and Max Borges — sits below 10 in terms of its score.
| Domain | Influence Score |
| properpropaganda.net | 31.5 |
| boltpr.com | 9.6 |
| 5wpr.com | 8.6 |
| avenuez.com | 6.7 |
| walkersands.com | 1.3 |
| maxborgesagency.com | 1.0 |
| edelman.com | 0.95 |
The citation consistency figure is arguably more interesting than the influence score. According to Scrunch, at 0.335, properpropaganda.net is the most consistently cited domain across our 695 tracked prompts. It is more consistently cited than LinkedIn, Reddit, or Forbes. That means AI systems have determined that our content is the most reliable reference text for the topics we write about.
There is one additional detail worth noting. Our site is being cited in AI responses that also mention other agencies like Avenue Z, Bolt PR, Edelman, Highwire, Hotwire, LaunchSquad, The Hoffman Agency, and Walker Sands. Effectively, properpropaganda.net is functioning as a reference text for the category, including in answers that discuss our direct competitors.
11. Takeaways
• Treat GEO as a slow channel that becomes a fast one. Our first signal of success took 3.5 months. By month six, the gains were compounding faster than we could have modelled.
• Build anchor hubs, not just service pages. Build pages that become resources for bots. Our owned site now has the highest citation consistency of any domain with meaningful volume in our tracked universe — higher than LinkedIn, Reddit, and Forbes. B2B companies, in industries where there is not a giant volume of media coverage in big outlets should pay particular attention to this tactic.
• Think of your site as a media outlet for bots. A meaningful portion of your content programme should be designed with AI systems as the primary audience. Structure beats style every time.
• Monitor agent traffic daily. Scrunch gives you a real-time view of what AI crawlers are reading. The feedback loop it creates is one of the most valuable operational tools in the programme.
• Category ownership shows up in the citation data. When your owned site is being cited in AI responses that also discuss your competitors, you have become the reference text for the industry. You want to be the source of explanation around how a category is framed.
• External authority matters more than ever. Podcasts, third-party articles, syndicated content, and citations are gold. You need a good PR team to be integrated into your GEO program.
• Founder voice is a GEO asset. A founder making consistent, authoritative arguments about a topic across multiple external platforms builds the kind of distributed signal that AI systems treat as validation.
• Own your category early. Once an AI engine learns from you, it keeps reinforcing you. The compounding effect is real and it accelerates over time.