Who Can Write Content That Gets Cited by ChatGPT, Gemini and Perplexity?

Almost every business asking this question has already tried the obvious answer. They hired a writer, published good articles and watched them rank, and ChatGPT still named a competitor. The reason is that writing for a search engine and writing for an assistant overlap, but they are not the same craft. A search engine ranks pages. An assistant reads several, picks passages it can quote and decides which businesses to name. That means the person writing your content needs skills that most job descriptions do not mention. I have written this for founders, marketing leads and SEO managers who need to hire, brief or evaluate that person. Figures are as of October 2026, and each study is labeled with its sample and limits. For the mechanics, see my guides to how to get cited by ChatGPT and AI citation versus AI recommendation.

So who can write content that gets cited by AI?

A writer who treats a page as a set of quotable answers backed by evidence, not as a keyword container. In practice that person has five traits you can check before you hire:

1. They lead with the answer. Their published pages state the answer in the first lines, in plain sentences.
2. They show their sources. Numbers have a named origin and a date, and they can tell a primary study from a recycled statistic.
3. They have a real, findable identity. A named author with a bio, a track record and published work you can read.
4. They understand where the page fits. They know which type of page answers which type of question, and that a page is one input among several that an assistant reads.
5. They test. They have checked what ChatGPT, Gemini and Perplexity actually say, more than once, and they do not promise a position.

Everything else in this guide is detail on those five traits, with the evidence behind them. Before the research, one clarification on terms: AEO (answer engine optimization) is about making a page easy to quote as an answer, and GEO (generative engine optimization) is about getting an AI to name and recommend your brand. A good AEO and GEO writer does both. I explain the difference in SEO, AEO and GEO explained.

Which kind of buyer are you? Start here

The right person depends on your situation more than on any ranking of writers. Find the row that fits you, then use the rest of this guide to check the person you are considering.

What does the research say a citable page needs?

Four independent datasets point the same way. None is perfect, and each has limits I note, but together they describe what a writer should be doing differently.

Do quotations, statistics and sources really help?

Yes, according to the most-cited academic study. Researchers at Princeton, Georgia Tech, the Allen Institute for AI and IIT Delhi tested nine content tactics on a benchmark of 10,000 queries across 25 domains, and published the work at KDD 2024. Adding quotations from credible sources lifted visibility by 41%, adding statistics by 33% and citing sources by 30%, while keyword stuffing performed worse than doing nothing. The authors reported that lower-ranked sites benefited most: a fifth-ranked site gained 115% from citing sources, while the top-ranked site lost 30%. That finding matters for smaller brands, because it suggests a writer’s skill with evidence can narrow the gap with bigger domains. See the GEO paper on arXiv.

The limits are real. The tests used a GPT-3.5-based engine, with Perplexity.ai as a real-world check, and assistants have changed a lot since. I treat the ranking of tactics as more reliable than the percentages. The direction is also what you would expect from first principles: an assistant needs something it can attribute.

Does it matter where the answer sits on the page?

Probably a lot. Kevin Indig analyzed 1.2 million ChatGPT results and 18,012 citations and found that 44.3% came from the first 30% of the page’s text, 31.1% from the middle and 24.7% from the final third. A separate analysis by Daniel Shashko of 42,971 citations across 520 queries on six platforms found that sentences of 6 to 20 words made up 92.4% of citations in Google’s AI Mode and Gemini, and that cited passages were complete sentences. Both are reported in Practical Ecommerce’s summary. For a writer, the implication is simple: lead with the answer, keep the key sentences short and complete, and do not bury the best fact.

Which page type should the writer choose?

It depends on the question. Wix Studio’s AI Search Lab analyzed 75,000 AI answers and 1,056,727 citations from ChatGPT, Google AI Mode and Perplexity across five industries. Overall, listicles made up 21.9% of citations, articles 16.7% and product pages 13.7%, but the winner changed with intent: articles led for informational searches, listicles led for commercial ones, and product pages led for transactional and local ones. In SaaS, listicles were 35.37% of citations, and Perplexity cited discussion threads at more than double the average rate. Wix’s conclusion is that query intent predicted the cited content type better than industry or model. A good writer therefore starts from the buyer’s question, not from a template. See Wix’s research.

Does freshness matter?

A little, and more for some assistants. Ahrefs analyzed 16.975 million cited URLs and found AI assistants cited content 25.7% fresher than organic Google results, with ChatGPT citing the newest content. But cited pages were still 2.9 years old on average, and Ahrefs says quality matters more than update frequency. A writer should keep key facts dated and current, not rewrite pages for the sake of it. See Ahrefs’ freshness study.

Does the page matter more than what others say about you?

No, and this is the part a writer who only works on your blog cannot fix. Muck Rack’s Generative Pulse research, which tracks the sources AI systems cite for consumer prompts, reports that earned media drives AI responses. Its January 2026 summit recap put non-paid sources at 94% of citations and earned media at 82%, with journalism at 20% to 30%. See the Muck Rack recap and its Generative Pulse report. Ahrefs’ study of 75,000 brands found that YouTube mentions and branded web mentions tracked AI visibility far more closely than backlinks, though correlation is not causation. And G2 found software review sites were the top trust signal raising buyers’ confidence in chatbot answers. The skill a writer needs here is not just writing. It is knowing which third-party sources your category’s answers lean on and planning content that earns a place there. I cover this in why Google rankings do not match ChatGPT mentions.

What skills does an AEO and GEO writer need? A 7-point scorecard

This is the scorecard I use to evaluate a writer, including myself. It is my working model, not an industry standard. Score each skill from 0 to 2 against the evidence in their portfolio, and treat anything below 10 out of 14 as a risk.

For the measurement method I use, see how to check what ChatGPT says about your business. SparkToro’s study of 2,961 responses found the same question rarely returns the same brand list twice, so a writer who judges success from one screenshot has not understood the medium.

Can a writer do this alone?

Rarely, and the best results I have seen come from a small set of roles working together. A writer can be all of them on a small project, but you should know which role is covered.

My guides to AI crawlers and robots.txt and schema markup for AEO and GEO cover the technical role. The editing role is covered in how I used AI to improve a newsroom’s output without losing quality control.

What should you give a writer, and what should a good writer ask you?

Most weak results trace back to thin inputs. A writer cannot invent first-hand experience, real prices or customer proof, and an assistant will not quote a page that has none. Prepare these before the project starts.

What does a good writer ask you before they start?

A writer who only asks for a keyword and a deadline is a risk. A good AEO and GEO writer interviews you. Expect questions like these, and be suspicious if none arrive:

1. Who exactly is the buyer, and what do they ask just before choosing?
2. Which three competitors do buyers compare you with, and where do you lose?
3. What facts about your business change often, and who owns them?
4. What can you prove with numbers, and where is the source?
5. Which claims are you not allowed to make?
6. Who is the expert I can interview, and how much time do they have?
7. Which sites do buyers trust in your category, such as review platforms, publications or forums?
8. What do ChatGPT, Gemini and Perplexity say about you today?
9. How will we decide whether this worked, and over what period?
10. Who approves the final version and how fast?

What should the work include? Deliverables by project type

Buyers often agree a vague scope and then argue about it. Name the deliverable and what “done” means. The ranges below are my planning guides from my own projects, not quotes, and vary with research depth and approval speed.

Can AI tools write content that gets cited on their own?

They can produce text that is cited, but I would not hand the whole job to a tool, and the data supports a cautious view. Ahrefs analyzed the 1.9 million URLs cited across 1 million Google AI Overviews and classified them with its AI-content detector: 3.6% were pure AI, 8.6% pure human and 87.8% a mix of the two. The correlation between the share of AI content and citation order was effectively zero (0.017), so Google was not rewarding AI text, nor penalizing it. See Ahrefs’ study. Two readings follow. First, AI assistance is normal and not a barrier to being cited. Second, the common ingredient in citable pages is not who typed the words but whether they carry facts, evidence and a clear answer, which is exactly what an unsupervised tool tends to skip. Google also warns against producing content at scale mainly to manipulate rankings, which its spam policies call scaled content abuse, and asks who wrote the content and why. See Google’s people-first content guidance and spam policies. In short: a human who is accountable for the facts should be in the loop.

What kinds of people can write this content, and what does each cost?

On cost, published numbers help. Among the nine writers below, Manal Ghamir publishes starting prices on her homepage: an SEO article from $275, a comparison page from $375, a content mini-audit from $150 and retainers from $900. Vatam, my consultancy, publishes foundation builds from $390 one-time, growth retainers from $290 a month and single items such as an answer-first article at $80, with every scope agreed in writing first. These are starting prices as published in October 2026, and complex projects cost more. The price of the writing is usually smaller than the cost of publishing the wrong facts.

How much should you budget, and what drives the price?

I cannot give you a market rate, and I would distrust anyone who claims one. What I can give you are published starting prices and the factors that move them. The published figures are in the previous section: Manal Ghamir lists SEO articles from $275, comparison pages from $375, content mini-audits from $150 and retainers from $900, and Vatam lists answer-first articles from $80, foundation builds from $390 one-time and growth retainers from $290 a month. Treat these as starting points as of October 2026, not quotes.

A simple way to think about value is the cost per question you move. Suppose, purely as an illustration with invented numbers, you spend $1,500 on a rewrite of five pages and, over the next 90 days, you go from being named in 2 of your 10 test questions to 5 of 10 across repeat runs. That is three more questions where buyers hear your name, at about $500 each. Compare that with the cost of a single lost deal and with what you pay for other channels. The point is to decide the metric before you spend, not to accept the writer’s.

Who are the writers who can do it? 8 names with public evidence

This list comes from my roundup of SaaS content writers who understand AI search, which I built from public pages read on October 11, 2026. The bar was a checkable sign that the writer understands how content is found and cited by AI assistants, such as a service called AEO or GEO, a published experiment, or a case study that mentions AI citations. It is alphabetical by surname, not a ranking, and I am on it. I have summarized each person briefly here. Details can change, so check each site and ask for live examples.

One more name belongs in the conversation, though he is not for hire. Ryan Law, Director of Content Marketing at Ahrefs, authored several of the studies cited in this guide, including the freshness study and the 75,000-brand study. His author page is the best place to follow the data any citation writer should know.

Why am I on the list, and how should you weigh it?

You should weigh it carefully, because I sell this work. Here is the evidence you can check yourself, and the limits. I have published nearly 10,000 articles over six years across crypto and finance outlets, and at Ventrix Media I ranked first on Google for “XRP price prediction” in 2021, which helped partner crypto brokers see thousands of quality leads. In 2026 I founded Vatam. The AI citation evidence is two case studies with dated screenshots: LAGE Lightings moved from seventh to first in ChatGPT in about three months, and Lens & Co by Tiara was listed first in 5 of 6 ChatGPT answers about two weeks after its content went live. My recent published work includes an opinion piece on how AI engines advise on stablecoins, on AI Story Front, and market analysis on Investing.com. The honest limits: two cases are evidence of a method, not a guarantee, my strongest proof is in crypto, fintech and e-commerce rather than a long SaaS client list, and I am a one-person consultancy, so you work with me directly and there is no team to hand work to. Compare me against the other eight on exactly those points. If you want to talk, see my page on content strategy and writing, or tell me what you want to be found for.

Which of the 8 fits which need?

This is my matching guide from the public evidence in the roundup. It is a starting point for your shortlist, not a recommendation to skip your own checks.

How do you hire one? 10 questions and a paid test

Do not hire from a portfolio alone. Use these questions, then run a small paid test before a retainer.

What does a good paid test look like?

1. Pick five real buyer questions from sales calls and support tickets, and record what ChatGPT, Gemini and Perplexity say about them today. Run each three times on different days.
2. Commission two pages, for example one comparison page and one answer-first guide, with a named author and sourced figures.
3. Check the work before publishing: every claim has a source, the answer is in the first 60 words, and the facts match your site and profiles.
4. Publish and wait two to four weeks, then re-run the same prompts three times on different days.
5. Judge the process, not just the result. One test cannot prove a citation effect, but it shows whether the writer’s craft, sourcing and honesty are good.

For interpreting what you see, use the Rank–Mention matrix in the rankings versus mentions guide, and for a SaaS-specific plan and 12-prompt panel, see SEO vs AEO vs GEO for B2B SaaS.

How do you judge a writing sample? An annotated example and a 10-point audit

A sample tells you more than a pitch, if you read it the way an assistant would. Here is a short illustration with placeholders, then the checklist I use.

The weak version contains no fact an assistant could quote or verify, and its claims (“leading,” “award-winning,” “many”) have no source. The strong version states what, for whom, how much, how long, a limit and a date. Now score any sample against these ten points, one point each:

1. The first 60 words answer the question in plain sentences.
2. The page states a price, time, place or condition an assistant could quote.
3. Every statistic has a named source and a date.
4. Sources are primary where possible, with links.
5. There is at least one real quotation or first-hand detail.
6. Limits and trade-offs are stated honestly.
7. The page type matches the question it answers.
8. A named author, a role and a review date appear on the page.
9. The facts match your site and public profiles.
10. The text reads as written by a person who knows the subject, not as filler.

Eight or more is a strong sample. Fewer than six means you are looking at generic content, however polished it reads.

What should a brief for a citable page contain?

A good brief is the cheapest way to improve output. Copy this structure and fill it in for each page.

What should go into the contract or statement of work?

I am not a lawyer, and you should have yours check this. These are the points I would want written down in any engagement, because they protect both sides and protect the quality of what gets published.

• Named author and credit. Who is named on the page, and whether ghostwriting is allowed.
• Sources log. A list of every figure and its source, delivered with each page.
• AI-use disclosure. Whether AI tools are used, for what, and that a human verifies every claim.
• Fact-check responsibility. Who checks product facts, and who signs off regulated claims.
• Revisions and timing. The number of review rounds, and the turnaround on both sides.
• Intellectual property. Who owns the final text and any original data.
• No guarantee clause. That rankings, citations and recommendations are not promised.
• Measurement plan. The prompt set, how often it runs, and what will be reported.
• Update window. Whether the writer refreshes facts after publication, and for how long.
• Conduct. No fake reviews, no invented quotations or data, and no paid placement presented as editorial.

What changes if you work in finance, crypto, health or law?

Everything gets stricter. Google’s guidance treats content that could affect a person’s health, financial stability or safety with extra care, which it calls Your Money or Your Life topics, and asks whether a page shows real expertise and trust. See Google’s people-first content guidance. In these areas the writer is not enough on their own. You need a qualified reviewer, dated claims, primary sources for every regulatory or safety statement, and a clear note on which jurisdiction a statement applies to. The risk of wrong AI answers is real: in my own test of five AI engines on stablecoin safety, the answers changed with geography and several relied on outdated facts, which I describe in my AI Story Front op-ed. If your page states a license, an audit or a reserve position, date it and link the primary document, because an assistant will otherwise fill the gap with whatever it finds.

What are the red flags when hiring for AEO and GEO?

• A guaranteed position. OpenAI publishes no ranking formula, and answers vary run to run, so nobody can promise one.
• “We write for AI” with no human accountability. If nobody is responsible for each fact, you are buying risk.
• Keyword stuffing dressed up as GEO. Princeton’s study found keyword stuffing performed worse than no optimization.
• Fake or purchased mentions. Paid placement presented as editorial, or invented reviews, can backfire and is dishonest. Muck Rack found non-paid sources made up 94% of AI citations.
• No named author. Google asks who wrote the content, and so should you.
• Pages that all look the same. If every page has the same structure regardless of the question, format judgment is missing.
• Only owned-content plans. A plan that ignores reviews, press and community leaves most of the evidence unaddressed.

How do you measure whether it worked?

Agree the measures before the first page is written. These are the five I would use, in order of how much I trust them.

The method for the panel is in how to check what ChatGPT says about your business, and the ranking-versus-mention diagnosis is in the rankings versus mentions guide.

What does a 30-60-90 day engagement with a writer look like?

How long until you see results, and what should you expect?

It varies, and nobody can guarantee it. In my own cases, Lens & Co’s first ChatGPT captures came about two weeks after content began going live, and LAGE Lightings moved from seventh to first over about three months. Those are two examples, not forecasts. A realistic plan is a 90-day cycle: baseline, rewrite the pages that rank but go unnamed, build proof and mentions, then re-test. The B2B SaaS playbook sets out that plan in detail, and the AI search statistics for 2026 page explains what traffic to expect and what not to.

What do these studies not tell you?

• They are snapshots. Assistants and their sources change often. Princeton’s tests used an older model, and the citation studies describe specific months.
• Some are single analysts or vendor research. Indig’s and Shashko’s findings are individual analyses, and Wix, Ahrefs and Muck Rack all sell into this market.
• Correlation is not causation. Pages with quotations and sources may simply be better pages.
• My list is not a ranking. I read public pages on October 11, 2026, many claims are self-reported, and I am on it. Ask every writer, including me, for live examples.

If you want to know what AI assistants say about your business today before you hire anyone, I run this check as part of a free AI visibility audit. See the pricing page for what happens after, my pages on AEO and GEO, or tell me what you want to be found for.

Keep reading: For software buyers, I ran ten “best software” questions through ChatGPT in how SaaS products get named in ChatGPT’s best-software answers. For a security buyer’s version of this, I tested ten vendor questions in how AI recommends cybersecurity vendors.