I usually hear it like this: “how do I get ChatGPT to recommend me?” That question leads nowhere, because it assumes there's a switch somewhere. There isn't. In this piece I gather the levers that actually move AI citations, in the order I use in audits, and I say plainly which ones are the strongest and which are just hygiene.
I'm writing about ChatGPT and Google's AI answers (AI Overviews, Gemini), because those make the most business sense today; that's where the real customers land. Perplexity, Copilot and Claude run on similar logic, so once you understand the mechanics once, you carry them over to the rest.
The wrong and right version of this question
The wrong version: “how do I hack ChatGPT?” It assumes a trick that doesn't exist. The better version is: why would a model that builds its answer from the whole internet cite me in particular? Once you switch to that second question, you stop hunting for a switch and start working on what actually decides it: trust and citability.
How LLMs choose what to cite
A model doesn't “rank links.” It builds an answer from many sources and picks the ones that meet three conditions at once: they're easy to cite (a clear, extractable point, not a wall of marketing), they come from sources it trusts (a recognisable brand, an author, a consistent entity), and they're corroborated in many places (it's not only your own site talking about your brand). Perplexity shows this most literally, citing sources right under the answer; ChatGPT and Google do it less visibly, but they choose the same way.
There's one more mechanism that's easy to forget: the model often looks at Google anyway. It turns your prompt into queries, pulls results from Google's index and grounds its answer on them. That's why classic SEO (indexing, rankings, clear structure) still matters a great deal here: if Google can't see you, the model has nothing to pull. And that's why there's no single “citation factor”: what counts is the whole network of trust signals, on your site and off it.
From practice: I've tested creating dedicated content for fairly niche queries, and it works, it can enter AI answers quickly. But the effect fades over time. That's the best proof of what this whole piece is about: a single move on your own site can work for a moment, but durable visibility comes from who the internet trusts over the long run, and from consistency, not a one-off.
The levers that actually move citations
I'll order them the way I order them in audits, from the biggest, most durable impact down to the hygiene without which nothing works. The order isn't accidental.
Off your site — this is where the biggest, most durable lever is
Digital PR and presence in trusted sources. The strongest signal isn't what you say about yourself; it's what trusted sites say about you. Coverage, mentions and citations on credible outlets build exactly the trust models look for. I treat this as work with a team and a budget, with transparent budgets run in the client's own name, not “buying links.”
Real users and UGC. Actual people who recommend your product and use your brand where the conversation happens (forums, communities, Reddit) carry a lot of weight now. Google and ChatGPT clearly lean on those sources. You can't fake it at scale; you need a product people actually want to talk about.
LinkedIn and personal branding of the right people. For many companies, especially in B2B, investing in organic visibility and activity on LinkedIn can do an enormous amount. Recognisable people from the company, their names, their takes, their presence, become part of the brand entity that models associate with you and cite.
Entity consistency. For a model to reference you, it has to know unambiguously who you are: a consistent name, sensible About and author pages, the same brand picture across many places. That's what ties all the signals above into one recognisable whole; more on that in AI visibility.
On your site — necessary, but less durable on its own
The SEO foundation. Since the model leans on Google's index, indexing and rankings still work for your citations. Without them, the rest hangs in mid-air.
Content freshness. A real lever, not decoration: updated, dated content gets picked more often. Material that just sits there eventually drops out of the answers.
Extractable answers and FAQs. A concise, unambiguous answer to a real question is a ready-made quote, the heart of AEO. A wall of marketing is not.
Technicals and structured data. Hygiene: access for AI bots (e.g. GPTBot), correct schema, clean structure. Necessary, but it's the floor, not a lever, which is why it sits lowest here.
And the most important part: notice that most of these levers don't sit on the SEO team's desk. That's why I don't think AI visibility should be built by SEO alone; it should be built by the whole inbound marketing team that owns the brand and its presence online. SEO, with its strong foundations, can lead it, but every brand activity across the web works toward AI recommendations. I've written more on which teams an SEO works with.
What not to do — and the biggest myth
A few things I see regularly that either hurt or waste budget:
- Schema stuffing. Piling on structured data with no backing in the content. The model still checks whether what you declare is actually on the page.
- Mass AI content with no value. Models pick sources they trust, not the biggest pile of text. A stack of thin pages works against you.
- Promising a specific outcome. If someone “guarantees a citation in ChatGPT,” just like guaranteeing rankings in Google, that's a red flag.
The biggest myth: that this is a one-off “AI optimisation.” It isn't. I've watched dedicated content climb into answers on a niche and then drop out over time. AI visibility is a state you hold with work: freshness, new trust signals, and presence where the conversation happens. It isn't a project with an end date.
How to audit and measure it
Start by asking the models themselves. Put your category, your recommendations, your comparisons to ChatGPT (with search) and Perplexity, and see who they cite and from which sources. Perplexity shows sources outright, so it's the fastest way to see who the internet in your niche trusts.
- Think in customer questions, not keywords. List the real questions people ask before buying, and check whether you show up in the answers.
- Track your share of citations over time. It's a trend, not a single measurement; watch whether it climbs month over month.
- Check the basics. Whether you're blocking AI bots, whether Google indexes you for those questions, whether the entity is consistent.
- Look off your own site. Whether you exist at all in credible external sources, in communities, on LinkedIn.
Then add tools. None is perfect, but they give you a picture. AI-visibility trackers like Chatbeat and Share of Model (and more entering the market) show your share of answers. GA4, a deep dive into Search Console, and Microsoft Bing Webmaster Tools can all hint at the prompts and questions you surface for. But you can infer a great deal manually: test different prompts, including in incognito on a clean account. It's not a perfect method, but even for a layperson it shows who lands in the answers and why. Then open those pages, get to know the brands, and work out why they were recommended.
Where to start this week
A few moves that give the most for the least — as long as you treat them as a quarterly rhythm, not a one-day list:
- Tidy up the entity — consistent About and author pages, the same brand name and description everywhere.
- Add concise answers and FAQs to your most important pages.
- Update and date your key content.
- List 3–5 credible places for digital PR and mentions.
- Nurture real reviews / UGC and the presence of the right people on LinkedIn.
- Check that Google indexes you for the key questions and that you're not blocking AI bots.
If you want this organised into one coherent strategy — that's what GEO / AI visibility is. And if the acronyms blur together, I untangle them in GEO vs AEO and in how to get cited in ChatGPT.
FAQ
Frequently asked questions
There's no switch. You build trust and citability — work spread over time and across many places, on your site and off it. Anyone promising a “ChatGPT hack” is selling certainty that doesn't exist.
Yes, a lot. The model often turns your prompt into queries and leans on Google's index, so without indexing and rankings there's nothing for it to pull. Classic SEO is the foundation AI visibility stands on.
ChatGPT and Google's AI answers (AI Overviews, Gemini) make the most commercial sense today — that's where the real customers are. Perplexity, Copilot and Claude follow similar logic, so you carry the mechanics over to the rest.
On niche queries the effect can be fast, but it fades without consistency. Durable visibility is a matter of months and ongoing work on trust signals — freshness, digital PR and presence where the conversation happens.
Yes, but treat it as a process, not a one-off. I've tested dedicated content for niche queries — it can enter AI answers quickly, but the effect decays without freshness and off-site signals.
Yes. Real users creating UGC and using your brand in communities like Reddit, plus personal branding of the right people on LinkedIn, are among the strongest trust signals models lean on today — especially in B2B.
