Last Updated on August 20, 2026 by Alex Birkett
I’ve got a few AEO case studies to show you.
AEO, or answer engine optimization, is the practicing of optimizing your brand’s presence for inclusion in AI responses.
It also often includes agentic or agent experience, which is optimizing products and websites for agentic tools.
Common metrics include AI driven traffic, share of voice in AI generated responses, citation share in AI search results, AI referred trials, or general AI discovery (self reported attribution and revenue).
These tend to coincide with SEO and general organic performance increases, so some weave in traffic growth as well.
I’ll cover the 5 answer engine optimization case studies I found useful or illuminating, and then I’ll walk through some caveats with the data and how to read case studies for maximum utility.
5 AEO Case Studies
1. Convert Grows LLM Visibility 81% and AI Citations 140% in 60 days
Think AEO results can’t come fast? Think again.
Obviously, the long arc of growing a brand and executing on impactful initiatives is the real winner’s podium. This game is one of cumulative advantage. Short term tradeoffs are rarely worth it.
However, as this case study from Omniscient Digital shows, a targeted approach combined with relentless execution can show up in the data fairly quickly, even in competitive and saturated spaces.
That is certainly true of experimentation and A/B testing platforms. Their client, Convert, was roughly in the middle of the pack in AI visibility, far from the leaders in the space.
Yet the team at Convert saw the opportunity in AI search earlier than their competitors and started executing on BOFU pages, structural content updates, technical performance, and off page brand mentions and signals.
The results: Over the course of 60 days, Omniscient grew Convert’s LLM prompt visibility from 31% to 55% across tracked queries. During the same time period, their competitors’ share of voice and AI visibility remains roughly static.
This was due to a cross functional approach where the Omniscient team worked with Convert to orchestrate resources towards key topic clusters and category entry points.
You can see the isolated effects of their website work in the impressive citation share growth over the same time period, jumping from 15% to 35%, a 140% growth rate.
Trina Moitra, Convert’s CMO, put it like this:
“Omniscient helped us understand how AI driven demand actually works and delivered replicable templates that we can apply to articles and BoFU landing pages, giving us a system we can continue to execute with confidence.”
2. An end to end AEO partnership with Zapier from strategy to attribution
Here’s a case study with almost no numbers in it. I’m including it on purpose.
Petra Labs’ write up of their end to end AEO partnership with Zapier is a process document, not a results document. It covers product level alias resolution for sub brands, statistically significant prompt sampling, a system of record for tracking initiatives, brand voice matching, PR outreach, and YouTube influencer strategy. The most interesting finding is that LLMs were confusing competitors’ documentation with Zapier’s own.
That’s a real problem with a real fix, and you can’t express it as a percentage.
Janine Anderson, Zapier’s Senior Web Marketing Manager, described the value this way:
“The ability to have somebody that works with us as a partner to figure out what to try and how to do AEO has been invaluable.”
Now here’s the part that should interest you more than the case study itself.
Profound also publishes a Zapier case study, which is no longer published on their blog, about the same program, and theirs has the numbers.
Roughly one in four Zapier signups was being influenced by LLM recommendations. 40% of homepage survey respondents attributed discovery to an LLM. Free trial signup rate from LLM influenced traffic ran 3x higher than Google organic. Citation share on competitor related prompts went up 4x.
3. Ramp Surpasses Competitors
Ramp is a spend management company that has seemingly worked with every platform and agency except my own. I digress.
Accounts payable automation was adjacent to their positioning, not central to it, and in AI answers they were nowhere. 19th in the category.
Between December 1, 2024 to February 15, 2025, Profound’s case study reports Ramp’s AI visibility going from 3.2% to 22.2%, roughly 7x, and their competitive rank moving from 19th to 8th in accounts payable.
The tactics are also named, which is rare.
Four pages: Accounts Payable Software for Small Businesses, Accounts Payable Software for Large Businesses, Top 6 Accounts Payable Automation Software, and AI in Accounts Payable.
Two of them generated 300+ citations in a single month. The insight behind them was that automation and AI content was getting cited in LLMs at a rate that traditional keyword research would never have flagged as a priority.
Ashley Nguyen, SEO Strategist at Ramp:
“Profound allowed us to uncover behavioral patterns that traditional SEO tools couldn’t fully capture.”
Here’s a detail I like from this one. Citation share went from 7.5% to 8.1%. That’s basically noise. It’s sitting right there next to the 7x visibility number, unhidden. If you were cherry picking, you’d cut it. Leaving a flat metric in the deck is a small act of honesty, and it’s a reliable tell.
Where it’s thin: Ramp passed “11 competitors,” but no one is named. Which is probably fine, and something due to liability and PR. Rank movement against an anonymous field is half a comparison. You know Ramp went up. You don’t know who came down, or whether the category simply got more crowded with mentions.
4.Novig Takes the Top Citation Spot
This is a great AEO case study that shows comparative data.
Novig is a prediction market sportsbook competing against operators with, in their words, a hundred times the marketing budget. Over six months, October 2025 to April 2026, Petra Labs’ case study reports AI visibility going from 0% to 21% across 1,500+ tracked prompts on ChatGPT, Claude, and Gemini.
The headline numbers are big: 27x AI-referred traffic, 35x organic, 10x direct. So in some ways, this is a rising tide lifts all boats – great execution across organic all around.
But take a look at the citation share table. In October 2025, novig.com sat 5th at 0.02%. The leader held 1.12%. By April 2026, Novig was 1st at 1.4%, and the former leader had fallen to 0.85%.
We can see here the relative movement that seems to have moved the needle, instead of just absolute increases from one company.
Citation share, at least in a stable comparison, is a bit of a finite pie. If your citation count goes up and nobody’s goes down, the most likely explanation is that the models started naming more brands per answer and everybody drifted up together. That’s why it’s a bit dubious when you hear “I got 3000+ citations.” Okay? Against what prompts? Does that matter? Rising tide, again.
A case study that shows a competitor’s decline is making a falsifiable claim about a fixed pie.
The timeline is disclosed too, which lets you sanity check the mechanism: prompt audit by week 2, first citations at week 6, top position by week 12, 21% visibility at week 24. Citations before visibility before traffic. That’s the order the causal chain should run in, and it’s a small thing to check that catches a surprising number of fabrications.
Nikhil Panu, Novig’s Head of Growth:
“Petra Labs gave us a presence in AI search that competitors with 100x the marketing budget still don’t have.”
5. Semrush
Okay, we’ve got Semrush dogfooding their own product here. Obviously we need a magnifying glass of scrutiny here!
However, the methodology here checks out and is actually quite useful.
Sergei Rogulin, Semrush’s Head of Organic & AI Visibility, published the internal program with three separate prompt sets and three separate results:
- 39 buying intent prompts: share of voice 13% → 32% in one month
- 726 prompt weighted set: ~15% → 25% over six months
- ~1,000 SEO topic prompts: 49% → 55% over six months
The tactics are also well documented in this one: product mentions embedded into pages that already ranked, content restructured into direct answers and tables, topic depth added where coverage was shallow, and off domain work on Reddit, LinkedIn, Medium, Wikipedia, and Quora. Tracked across ChatGPT, Google AI Mode, Perplexity, and AI Overviews.
Nothing crazy, right? Good old fashioned organic marketing execution.
But the reason this one earns its place is the caveats, which Rogulin volunteers rather than buries:
“We can’t fully isolate which tactic drove what, since we ran them at once.”
Bingo.
And on revenue: separating AI’s impact from paid search and other channels proved difficult. The data is improving, but they’re not there yet.
That’s a vendor, in a piece of marketing for their own product, telling you their attribution doesn’t work. When someone with every incentive to overclaim underclaims instead, pay attention.
The Problem with AEO Case Studies
I know I just listed several AEO case studies, but here’s the section where I warn you to take anything you read with a grain of salt.
First off, any case study in any space is published with incentive alignment towards making the brand, agency, platform, or vendor look awesome. There’s a survivor bias.
I learned this early in my career while working in conversion rate optimization and experimentation. Without context in this case, sample size, duration, conversion lift, test protocol it’s difficult to say whether the win was real or illusory. Rondeau wrote a great piece on this years back.
With AEO, it’s even harder. Here’s why:
There’s no standardized set of prompts, citations, or metrics to compare apples to apples.
When you read a case study about citation lift, you have to ask, “against which prompts?” Same goes for AI visibility.
There’s a good faith interpretation here if it’s an agency or a platform, presumably the client was happy enough to put their name behind it. Anonymized case studies are a whole other story…take those with a massive grain of salt if not fully disregarding them.
Ultimately, what matters most is pipeline, or in absence, comparative share of voice against competitors. It could be the case, for instance, that model changes impacted brand visibility for all top brands, so everyone in the pack could rise by 5%, which is meaningless if it’s simply the case that more brands are mentioned in answers.
You see what I mean here? There are so many confounding variables.
That’s why I tried to represent AI search case studies above that have comparative share of voice and named brands.
There’s another problem: any AEO case study that reports on AI referral traffic or conversions from AI traffic is dealing with two huge, ruinous confounding variables
- Demand growth
- Mode changes
On the first, put simply, everyone had case study worthy growth last year. Why? More people used AI tools, so more people clicked through on the citations, and everyone’s AI traffic grew. Didn’t matter what you implemented or changed. Rising tide lifts all boats. Ignore these case studies for the most part, especially if you don’t get date ranges.
Second point is tougher, but my agency has access to several dozen B2B accounts so we see aggregate patterns. Some months, ChatGPT spikes. Turns out, it’s because they are linking out more prominently. Some months, AI overviews shrink. Turns out they are mentioning fewer brands.
In aggregate, a single platform changing their model or interface could alter your aggregate metric, even if the rest of your data is trending in a different direction.
Thus, it’s absolutely easy to cherry pick data.
As always, caveat emptor.
How to Read an AEO Case Study
Here are instructions for being a discerning reader, knowing what you can learn and what to disregard as pure marketing fluff or disingenuous cherry picking.
Look for contextual cues, comparisons, and counterfactuals.
Consider the denominator, or in other words, why this metric is meaningful in context.
Every AI visibility number is proportioned against a self-selected set of prompts, and the case study is under no obligation to tell you what’s on the bottom of that selection. Semrush’s 13% → 32% and its 15% → 25% describe the same six months of work. The only difference is 39 prompts versus 726. If a case study won’t tell you how many prompts, which models, and over what window, it’s difficult to discern the true impact to the business.
Consider the comparisons. Share of voice and citation share are near enough to zero sum that somebody has to give ground for you to gain it. When a case study shows a client rising and says nothing about the field, it’s very possible the field rose too.
Consider the counterfactual. What happened to Google traffic in the same window? Are we counting attribution instead of incremental performance? What would have happened if we did nothing? In many of the popular answer engine optimization stories, we see AI traffic reported, but absent of context, it’s hard to know if that was good, bad, or just the natural course of consumer behavior changes. That’s why controlled experiments are so excellent (and I’d like to see more quasi experiments in this space).
Audit the logic, the process, and the inputs, not just the outputs
The output is so specific to the company, prompts, and industry at hand. It doesn’t really matter if they increased by 20%, 50%, 2000%. You can’t pull that into your own context. The reasoning, the process, the work, and the experiments are what you can learn from.
I thought Ramp’s case study where they actually spoke about the four pages that mattered was insightful.
Semrush documenting why it migrated from 39 prompts to 726 is useful, too, especially in designing the logic of prompt selection and measurement,
Petra Labs finding that LLMs were serving competitors’ documentation as Zapier’s is worth more than a lift chart, because it’s a failure mode you can go check for on your own brand this afternoon.
Interrogate the inputs, especially the ones that go unmentioned.
One AEO case study I read reports strong growth in non-branded citation share; further down, the tactics include sponsored placements on third party review sites.
This is a legitimate play, of course. But if you’re a marketing leader, it’s really useful to know that’s what it took. There are a million and one ways to “get brand mentions.” Pay, borrow, beg, create. It just changes how replicable it is based on the inputs that went into it.
Helpful also to ask what the client already had. Domain authority, an existing Reddit presence, a PR budget, a 40-person content team. The starting position determines how much of the result you can expect to reproduce. No shade to the mega brands out there, but I highly doubt schema has anything to do with Nike’s AEO success. They are winning because they are Nike.
Understand that this space is new, volatile, and that experimentation is a key mental model in AEO
Every AEO case study you will read is an observational study, a case report where the data presented was selected by the person publishing it, usually no control group, conducted on a measurement instrument that the vendor is in charge of.
That’s just the reality. It’s not unique to AEO or AI systems (though challenges are especially strong in this space).
Perhaps I am a cynic. But perhaps we should all be a bit more discerning.
I don’t throw them out.
I use them a way a researcher would use a case report. A smile and a nod and a “huh, that’s interesting. I’d like to research that further.” In other words, to generate a hypothesis to test.
Also, this stuff is volatile. Unlike traditional search engines, AI powered search has unlimited variables, probabilistic outputs, and a high degree of personalization and memory. Scoreboard changes often.
Principles, generally, are robust. But I’d keep an open, growth mindset here.
Which points at the only real answer: run your own tests.
Baseline data, hypotheses, and as much structure in your experiment design as the AI systems allow.
Read the case studies for plays worth testing. Get your proof from your own instrumentation.
Conclusion
None of the case studies above are proof.
I picked them because they’re useful for learning and generally intellectually honest in a vast sea of fluff and grift.
Hopefully this was inspiring or educational. I would highly recommend, in architecting our AEO strategy, to use these as minor data points, mostly inspiration, and spend more time thinking for yourself.
We’re early enough that nobody has this figured out, including the people selling it, including me.
The category will standardize eventually, in terms of some shared definition of visibility, some agreed prompt set methodology, some norm about disclosing date ranges. It’s not there.
So hey, think for yourself, have fun experimenting, and get a little creative.