Why You Shouldn't Trust Anyone Promising "Guaranteed" AI Visibility (And What Actually Works)
In 30 seconds
If someone promises they can guarantee your brand appears in ChatGPT responses, you should be sceptical. Very sceptical.
AI visibility matters, absolutely. But the mechanics of how large language models work make those specific promises impossible to deliver with certainty.
Here's what marketing directors need to know about AI visibility, why direct attribution claims are a red flag, and what credible agencies focus on instead.
Welcome to 2026, where AI dominates every marketing conversation. And will do for a while yet, I’m afraid.
In today’s massive LinkedIn AI shouting pit, amongst the “SEO is dead” mourners and “Ads are coming to GPT” flagellators, you'll find a stream of voices promising temptingly definitive answers. And it’s not surprising. After all, certainty sells. When everything's up in the air, guarantees feel like solid ground.
But if you want to trust what your marketer is saying, you need someone comfortable with ambiguity. Definitive promises are actually red flags. Fool's gold.
AI isn't set in stone (and the AI companies themselves will tell you that). So, what can you do that will genuinely help your brand? And what kind of marketing waffle should actually have you running for the hills? (I’m not being paid per idiom, I promise).
You're in the right place to find out.
The problem: AI Is a black box
You can't step in the same AI river twice (sorry).
Recent research from SparkToro found there's less than a 1 in 100 chance that ChatGPT or Google's AI will give you the same list of brands in any two responses, even when asked the identical question.
Beyond that, unlike Google Search (which provides Search Console data showing exactly what people typed and where they clicked), ChatGPT is fundamentally opaque. OpenAI doesn't provide a dashboard for brands to see how many times they were mentioned.
In short, you have:
No analytics. There's no way to track how often your brand appears in responses.
Extreme variation. Nearly every response is unique in the list of brands presented, the order of recommendations, and the number of items on that list.
Anyone promising "we can prove exactly 50 people found you through ChatGPT this month" should raise your heckles. With current technology, that level of tracking precision is impossible. If someone's promising it, they're guessing at best, fabricating at worst.
Us when we see those “rank number 1 on Chat GPT” adverts 🚩🚩🚩
The alternative: measuring Share of Model
So if direct attribution is impossible, what should credible agencies measure instead? The answer is Share of Model, or visibility percentage. It’s a more honest approach to understanding AI visibility.
Jem Leslie, senior search & content strategist explains, “Rather than promising clicks or rankings, we should focus on increasing the probability that AI systems mention you.”
A sophisticated agency does this by influencing the training data with your brand's expertise: getting you into top-tier news sites, creating moments that influence Reddit discussions, niche blogs, and industry reports (the sources that AI models scrape and learn from).
When SparkToro ran the same prompt 60-100 times, certain brands consistently appeared more frequently than others. This gives an idea of the percentage at which a brand appears in responses. Measured across dozens or hundreds of queries, this provides a reasonable indicator of how prominent the brand is within the AI's consideration set.
The measurement approach uses third-party AI tracking tools to sample queries repeatedly and calculate what percentage of responses include your brand. It shows relative visibility, rather than a hard metric promising specific placement. It’s way more honest than claiming direct attribution or guaranteed rankings that simply don't exist.
What you can measure
While you can't reliably track mentions in pure conversational AI like ChatGPT, you can track citations in Search-Generative Engines (SGE) like Google's AI Overviews. These provide source links alongside their AI-generated answers.
Jem explains, “Think of your performance in AI Overviews as an early indicator for how you'll show up in ChatGPT. Both systems work on similar principles, but they have different timescales. AI Overviews only use live data. ChatGPT relies on training data that only updates periodically – as of Feb 2026, its knowledge base cuts off at August 2025. When ChatGPT performs RAG (pulling information live from the web), it starts behaving more like an SGE. So, tracking your SGE visibility gives you a good picture of your AI presence, both at the live retrieval stage, and in the eventual static training data, when it gets a refresh."
This is where editorial digital PR becomes even more invaluable. If ChatGPT cites a BBC or Wired article to answer a prompt, and you're the featured expert in that article, you've effectively "won" the AI placement. Your expertise has been endorsed by a publication that AI models trust as a citation source.
The logic is goes like this: AI systems learn from authoritative sources. If your brand consistently appears as the expert voice in publications like The Guardian, Forbes, or industry-leading trade media, AI models are more likely to reference you when answering related queries.
You can't control every instance, but you can increase the probability.
What this means for your agency choice
The SparkToro research uncovered another critical insight: the massive diversity in how people ask AI tools questions.
When researchers asked 142 people to create prompts about the same topic (choosing headphones for a travelling family member), barely two prompts looked similar. “what are the best headphones with noise cancellation and long battery life?” vs “which headphones should I buy for my mum who travels a lot?”. People don't reduce their queries to simple keywords like they do in Google. They get creative, specific, and weird. They sound like people again.
Why does this matter? Your brand needs to be associated with the core concept, rather than just specific keyword phrases. Those two headphone prompts are semantically different, but if your brand has sufficient authority in that space, it should appear in both responses.
There’s more data that backs this up: in 2024, search queries triggering AI Overviews grew from an average of 3.1 words in June to 4.2 words by December. Searches with 5 words or more have doubled since ChatGPT launched, with an October 2025 study finding more than three quarters of searches are 5 words or longer.
So, it looks like the more people use AI, the weirder their prompts become. That behaviour is still evolving.
It means agencies focused on keyword optimisation for AI are missing the point – because keywords are OUT in ChatGPT, and also because AI models don't work like traditional search engines. They're trained on the broader corpus of content about your industry, and they surface brands based on how frequently and authoritatively those brands appear. You can't game this with technical tricks. You need genuine editorial presence.
Let’s play “Red Flags vs Green Flags”
When evaluating AI visibility services, watch for these patterns:
| Red Flags | Green Flags |
|---|---|
|
Promises a top 3 spot in ChatGPT or Gemini 🚩 These systems don't have ranked lists like Google Search. Plus, the SparkToro research found less than 1 in 1,000 chance of getting the same ranking order twice. |
Focuses on visibility percentage, not rankings 🟩 Talks about increasing the percentage of responses where you appear, measured across dozens or hundreds of queries. Acknowledges rankings are essentially random. |
|
Proprietary methodologies that package basic PR as revolutionary 🚩 Launches frameworks that repackage fundamental PR principles (create content, secure coverage, build authority) as proprietary breakthroughs. The SparkToro research proves there can't be a secret method. |
Transparent about what they're actually doing 🟩 Explains their approach in plain language: securing coverage in publications AI models cite, building thought leadership, creating content that answers real questions. No jargon, no mystification. |
|
Academic endorsement without practical research 🚩 Uses endorsements from business school professors who study marketing theory, not practitioners who test approaches against real AI systems. Academic credentials do not equal empirical validation. |
References independent, transparent research 🟩 Cites research from independent sources that openly share methodology and raw data, like SparkToro's 2,961 AI responses analysed. Transparent, peer-reviewable work. |
|
Claims a "secret hack" or "direct line" to OpenAI 🚩 Suggests special access, technical AI tagging, or hidden code that tricks models into favouring your brand. There is no backdoor. This preys on marketers' lack of technical knowledge. |
Explains results are probabilistic 🟩 Acknowledges they're increasing the chance of mentions, not guaranteeing them. Talks about influence, not control. Reflects genuine understanding of how LLMs work. |
|
Vague language that reads like nothingburger jargon 🚩 Uses phrases that sound strategic but mean nothing specific: "AI-native frameworks", "leveraging synergies". Designed to make you nod along without understanding what you're buying. |
Uses E-E-A-T as the main strategy 🟩 Focuses on Experience, Expertise, Authoritativeness, Trust. Building genuine authority through consistent, high-quality content across reputable platforms. Not sexy, but it works. |
What Bottle does differently
Our approach to AI visibility is built on editorial digital PR: getting clients featured in the publications that AI systems cite as authoritative sources. We don't promise we can control ChatGPT. We promise we can increase the likelihood it references you by building the kind of brand authority that AI models learn from.
When we place a client in the BBC, The Guardian, or Wired as the expert voice on their topic, we're creating a citation source that influences what AI systems recommend. When ChatGPT or Perplexity answers questions in that space, they're more likely to surface articles where our clients are the featured authorities. We're generating backlinks for SEO, yes, but also building the editorial presence that AI systems learn from.
We also understand that AI systems increasingly pull from Reddit discussions, YouTube videos, and social media platforms, rather than just traditional news sites. Our approach accounts for this. We build stories that can translate to presence across the entire ecosystem where AI models learn, from earned media coverage to owned social content.
You're maximising the footprint of your genuine expertise, rather than gaming the system.
Are you a marketing director?
Read this.
AI visibility matters. As consumers increasingly turn to ChatGPT, Perplexity and Google's AI Overviews instead of traditional search, brands that don't appear in these systems will lose ground. Research from Pew shows that users increasingly skip traditional link lists in favour of AI-generated summaries in Google. Your brand either becomes part of the conversation that AI learns from, or gets left out.
That doesn't mean you should accept misleading promises from agencies that claim they can guarantee specific outcomes in systems they fundamentally can't control. The agencies worth hiring are the ones who understand the limitations, focus on what works, and build strategies around influence rather than false certainty.
The uncomfortable reality: an agency cannot control ChatGPT. But an agency can influence the sources that ChatGPT reads.
For more detail on the inconsistency of AI responses and what it means for tracking, SparkToro's full research report provides extensive data and methodology that every marketing director evaluating AI visibility services should read.
Want to talk about how digital PR can build your brand’s authority across the sources AI systems learn from? Get in touch.