How to get cited by ChatGPT and Perplexity (2026)

For years the goal was a blue link on page one. Now a growing share of people ask ChatGPT or Perplexity a question and read a synthesised answer that cites a handful of sources — and never scrolls a results page at all. If you run a one-person business, that shifts the question from “how do I rank?” to “how do I become one of the few things the machine quotes?” This is the honest version: how those citations get chosen, what a solo can actually influence, and why nobody — including the people selling courses about it — fully controls the outcome.

Two different machines: training vs retrieval

The single most useful distinction is the one most posts skip. There are two ways your content might end up inside an AI answer, and they behave completely differently.

Training data is the frozen snapshot the model learned from when it was built. It lags reality by months, you can’t edit your way into it, and you certainly can’t guarantee the model “remembers” your brand. Chasing it is mostly a waste of a solo’s time.

Live retrieval is what an answer engine fetches right now to answer a current question — it searches a web index, pulls a few candidate pages, and synthesises an answer with citations. This is the part you can influence today. A clean, indexable, accurate page can be retrieved and cited this week, regardless of when the underlying model was trained. So the entire practical game is: optimise for retrieval, not for being memorised. That mindset shift is the foundation of generative engine optimization (GEO), and it’s where it quietly diverges from classic SEO.

How a citation actually gets chosen

Strip away the mystique and the pipeline looks familiar. The engine retrieves candidate pages for the query (often using a real search index underneath), ranks them by relevance and apparent trustworthiness, then writes an answer and attaches a few sources to back specific claims. Two things matter more than any clever trick:

• Retrievability. If your page can’t be crawled, parsed and matched to the question, none of the rest applies. This is ordinary technical hygiene — indexable, fast, clearly written — and it’s the same baseline that gets a new site any AI traffic at all.
• Reputation, broadly read. Engines lean toward sources that look corroborated. A claim that appears consistently on your site and is echoed elsewhere — communities, directories, other people’s articles — reads as more reliable than one that exists only on a page you control. That’s the part solos under-invest in, and it’s the part that compounds.

You’re not gaming a ranking. You’re trying to be the obvious, well-supported answer the machine would be slightly embarrassed not to cite.

Be the clear, single answer to one question

Answer engines reward content that resolves a specific question cleanly, not pages that wander. For a solo with limited hours, that’s good news — focus beats breadth.

What “clear” looks like in practice:

• One page, one question, answered up top. Lead with the direct answer in a sentence or two, then support it. Don’t bury the conclusion under 600 words of preamble; the engine (and the reader) wants the answer, then the reasoning.
• Factual, specific, consistent claims. Concrete statements — figures, definitions, steps — are easier to retrieve and quote than vague prose. Crucially, keep your claims consistent across your own site. Contradicting yourself page to page makes you a worse citation candidate.
• Structure the machine can read. Real headings, short paragraphs, lists, plain question-and-answer blocks, sensible FAQ. This isn’t a magic schema trick; it’s making the answer trivially extractable.

Presence beyond your own site

This is the lever most one-person businesses ignore, and it’s the one I’d prioritise. Being the authoritative answer isn’t only about your page — it’s about how often a credible, consistent version of your claim shows up across the web, where the engines also look.

Honest, non-spammy moves a solo can actually do:

• Get genuinely referenced in places that get cited — a useful answer in a relevant community, a guest piece, an entry in a credible directory or comparison, a tool or dataset others link to.
• Keep the facts identical everywhere. Same numbers, same definitions, same name, across your site, your profiles and anywhere you’re quoted. Consistency is what makes a claim look corroborated rather than coincidental.
• Earn mentions, don’t manufacture them. A handful of real references on trusted sites beats a pile of low-quality ones, which the engines (and search) increasingly discount or penalise. Slow and credible wins.

There’s a strategic edge here for indie makers: as answer engines absorb the “which is best?” query, they’re quietly hollowing out generic comparison sites. The defensible position is being the primary, trustworthy source on a narrow topic — exactly what a focused solo can be and a thin affiliate aggregator can’t.

The small technical bits (and what not to over-invest in)

A few low-effort hygiene items help; none is a growth hack, and it’s worth being clear about that so you don’t burn a weekend on the wrong thing.

• llms.txt is a tidy, machine-readable map of your key pages. It may make a well-built site easier to parse, but adoption is uneven and it is not a ranking lever — here’s the realistic take on llms.txt. Add it in five minutes, then forget it.
• Don’t accidentally block the crawlers. Check that your robots rules and any bot-blocking don’t wall off the AI crawlers you actually want retrieving you. This is the most common own-goal.
• Keep pages fresh. Retrieval favours current information. Updating the date and the facts on a page that’s still accurate is worth more than spinning up new thin ones.

That’s roughly the entire technical list. The leverage is in the content and the off-site presence, not in a config file.

How to measure it (honestly)

Don’t look in Search Console — AI citations mostly aren’t classic search clicks, so they won’t show up there. The trail is in referral traffic: visits arriving from chatgpt.com, perplexity.ai and similar. You usually have to set those segments up deliberately, because default reports under-count or mislabel them — here’s how to track AI/ChatGPT traffic in GA4.

And the caveats, because anything else would be dishonest: a large share of AI answers never click through at all, so referrals undercount your real influence; attribution is patchy and noisy; and the numbers swing as engines change behaviour. Watch the trend over weeks, treat any single reading as one data point, and don’t build your business plan on a metric this young.

The takeaway

• Optimise for retrieval, not training. The frozen model snapshot is out of reach; the live-fetched page is the part you can actually get cited from today.
• Be the clear, factual, single answer to a specific question — lead with the answer, keep claims concrete and consistent across your site.
• Presence beyond your own site is the under-used lever: credible, consistent mentions elsewhere make your claim look corroborated, which is what earns the citation.
• Technical bits are hygiene, not magic — add llms.txt, don’t block the crawlers, keep pages fresh, then move on.
• Measure with referrals, not Search Console — and stay honest that it’s noisy, under-counts, and changes fast. There is no guaranteed formula, only better odds.

Part of the complete SEO for solopreneurs guide.