Publishing the same content and expecting to show up on ChatGPT, Gemini, Perplexity and Claude in the same way is like running a single ad and hoping it works the same on TV, radio, newspapers and social media. Every AI platform retrieves information from different sources and with different logic — and whoever produces a strategy tailored to each one occupies the space that those using a one-size-fits-all approach always leave empty.
Apple Intelligence on the iPhone is the most intimate of AIs: it runs on the device, it guarantees you absolute privacy, it answers without sending your data elsewhere. But the model that answers you is limited by the data it was trained on. If you’re not inside Apple’s training, you’re simply invisible to the most loyal Apple user — the one who opens the evolved Siri instead of the browser.
The same applies, in mirror image, to ChatGPT, Gemini, Perplexity, Claude and Copilot. Each of these AIs has a different knowledge source and a different answering mechanism. A “same for everyone” strategy makes you appear on a single one, maybe the wrong one for your customer. In this article I explain why you need a dedicated plan for each platform and how to build it without multiplying the work by five.
What I mean by a platform-specific strategy
For years in classic SEO we debated whether it was worth optimizing for Bing as well as Google. The answer was: not much, because Bing accounted for 5% of traffic and the ranking signals were similar. With generative AI the math changes: ChatGPT has 200+ million weekly users, Perplexity is growing on informational queries, Apple Intelligence will reach the devices of hundreds of millions of users. These aren’t five versions of the same engine. They are five different engines that draw from different sources.
The underlying principle is simple: if the knowledge source changes, the point at which you make yourself visible has to change too. In the previous articles of this series I told you about tokenization and E-E-A-T for AI: those mechanisms hold for all models, but how you bring them to the right platform changes from one to the next.
The platform-by-action matrix
What we have is the companies’ public documentation and the observations of those who monitor AI citations every day. From these two sources I derive an operational matrix that I use with my clients.
ChatGPT has two souls: the base model with a training cutoff and the search mode that queries Bing in real time. To show up in the first one you have to have been indexed by OpenAI’s crawlers before the cutoff: older, authoritative content, cited elsewhere. To show up in the second you have to perform well on Bing.
Gemini is the most tied to the Google ecosystem: it draws from the Knowledge Graph, from structured data and from the Google index. If your brand doesn’t have a KG entry and a solid Organization schema, Gemini struggles to cite you. You can check it with Google’s Rich Results Test.
Perplexity rewards freshness and crawlability. It needs up-to-date, accessible content with clean internal citations. If your robots.txt blocks AI crawlers or your pages are behind authentication, you’re cut out. You can check this with the robots.txt tester.
Claude favors long, well-structured content with articulate reasoning. Anthropic’s training data includes many editorial and academic corpora. For short, promotional articles, Claude rarely cites you.
Microsoft’s Copilot is effectively Bing on steroids: whoever is well positioned on Bing Webmaster Tools starts with an advantage.
Apple Intelligence is the most closed case: it runs on-device, a compact model, Apple training. You can’t optimize in real time. Only the authority accumulated on the corpora Apple has chosen counts.
Publishing 30 mediocre articles to show up everywhere works worse than publishing 8 done well.
The observation I made on producers of Vino Nobile di Montepulciano
Here I get concrete. Since the start of 2025 I’ve been monitoring eight producers of Vino Nobile di Montepulciano DOCG — an appellation produced exclusively in the municipality of Montepulciano, in the province of Siena, one of the most recognized in Tuscany. I chose this sector because it’s a closed perimeter (a DOCG appellation with a regulatory specification and registered producers) and because it’s a typical use case for consumer AI: the user asks for advice on the “best vino nobile to pair with a Florentine T-bone steak” or “wineries to visit in Montepulciano.”
I compared the answers of Apple Intelligence in beta (on an iPhone with iOS 18.x) and ChatGPT with search enabled across 24 rotating queries for eight wineries, from February to October 2025. An indicative test, not a scientific study: small sample, queries not balanced for intent, results that vary depending on the prompt. But the pattern is clear enough to be useful.
On Apple Intelligence, two wineries out of eight appeared recurrently in the answers; the other six were never cited by name or were cited generically (“a historic winery in the area”). On ChatGPT with search, seven wineries out of eight appeared at least once across 24 queries, with different frequencies. The difference between the two platforms was enormous: those who were invisible on Apple appeared regularly on ChatGPT, and vice versa in one case out of two.
From this follows an operational consequence: if you’re a Vino Nobile winery and you invest only to show up on one platform, you’re accepting being invisible to half or two-thirds of AI users in the medium term. Getting the result here means multiplying your points of presence, not picking the “winner.”
Check your robots.txt with the tester: make sure GPTBot, ClaudeBot, PerplexityBot, Google-Extended aren’t blocked by mistake.
How to build the matrix without multiplying the work
The most common mistake I see is thinking you need five sites, five blogs, five editorial strategies. That’s not how it works. You need one central asset (site + pillar content) and five different triggers that make it visible to each platform.
For ChatGPT: publish content that gets cited by authoritative sources (publications, industry magazines, Wikipedia). The training data updates slowly: get ahead of it. In parallel, optimize for Bing with Bing Webmaster Tools to intercept search mode.
For Gemini: tend to your presence on Wikidata and Google Business Profile, add Organization and Product schema to the site, monitor impressions in Google Search Console. For the KG I’ll point you to the article on the Google Knowledge Graph.
For Perplexity: publish frequently, make sure robots.txt doesn’t block the declared AI crawlers, keep clean internal links between the hub and satellite content.
For Claude: write long-form (1500+ words) with structured reasoning, cited sources, recognizable authors. On your authors’ content, Author Entity Recognition matters.
For Copilot and Apple Intelligence: the first follows Bing, the second is the slowest and depends on accumulated authority. For Apple there’s no direct technical lever: what counts is the accumulation of citations on editorial corpora over the medium to long term.
The mistakes I notice most often
Treating ChatGPT as the only target. It happens because it’s the best-known name, but if your customer uses Apple or Google, you’re optimizing for the wrong platform. The result: you’re happy with your internal test and invisible to half the market.
Confusing Copilot with ChatGPT. They’re two different things: Microsoft Copilot draws from Bing, ChatGPT in base mode draws from the training. Same company, OpenAI, but different flows and sources.
Blocking AI crawlers to “protect your content”. I see SMEs with robots.txt that exclude GPTBot, ClaudeBot, PerplexityBot. The result isn’t protecting the content: it’s making it invisible to the generation that’s replacing search.
Thinking it’s enough to publish more. Frequency helps Perplexity, not Apple. Publishing 30 mediocre articles to show up everywhere works worse than publishing 8 done well.
Here’s what to do in the next 30 days
- Open the five main AIs and run the same 10 queries about your sector (e.g. “best producer of Vino Nobile in Montepulciano”). Note where you appear and where you don’t.
- Check robots.txt with the tester: make sure GPTBot, ClaudeBot, PerplexityBot, Google-Extended aren’t blocked by mistake.
- Check the Organization schema on the Rich Results Test and complete your Google Business Profile.
- For every platform where you don’t appear, identify a probable cause (training data absent, schema missing, robots blocked, content too short) and a single action.
This is an entry-level audit: it tells you whether you’re above or below the minimum threshold. The real analysis, with continuous monitoring of AI citations by sector and benchmarking against competitors, requires professional tools and dedicated time.
Visibility in AI answers is plural
“Being visible in AI” in the singular no longer exists. What exists is being visible in ChatGPT, Gemini, Perplexity, Claude, Copilot, Apple Intelligence — and each one calls for a different trigger on the same central content. A platform-specific strategy doesn’t mean five strategies: it means one strategy with five points of application.
In the upcoming articles of this series I go into platform-by-platform detail: how to optimize for Copilot inside Microsoft 365, how to work on freshness for Perplexity, how to build the Knowledge Graph presence that Gemini rewards. If you want a reference on the authority that runs across all platforms, I’ll point you to the article on implicit reference weight.
