Get cited in AI answers
Becoming the source an answer is built from, not just a name in it
Nobody controls whether a model cites anyone, and we will not pretend otherwise. What can be changed is whether you are the kind of source worth drawing on: the origin of a fact rather than a repeater of one, stating claims in a form that can be lifted cleanly, dated, maintained, and unambiguously attached to your organisation.
Why the answer quotes somebody else
Four reasons a competent expert never gets quoted
This page assumes you are already present and described accurately. The question is narrower and more irritating: why is the material being drawn on always somebody else’s?
- You are repeating facts rather than producing them.
- Your best pages quote an industry figure that originated at a trade body, a vendor study or a consultancy report. That makes your page a route to the original rather than a source in its own right, and anything drawing on it has an obvious reason to skip a step. No structural change to the page repairs this, because the problem is upstream of the writing.
- Your expertise is real and unliftable.
- The knowledge is genuine and it is spread across two thousand words with the subject in one paragraph, the number in another and the caveat in a third. Nothing can be quoted without reassembling it, and reassembly is paraphrase, which arrives without your name attached. This is the most common failure among the businesses that most deserve to be cited.
- Nothing on the page says when or who.
- No publication date, no review date, no named author, no method. A claim without a period attached cannot be evaluated as current, and a claim without a person behind it cannot be weighed for expertise. These are small omissions that make an otherwise excellent page structurally weaker than a mediocre one that includes them.
- The claim is not clearly yours.
- The material appears on a page with no clear organisational identity, or it is syndicated in four places with different framings, or it sits under a byline that does not resolve to anyone. When the same content exists in several forms with no obvious origin, whatever draws on it has no reason to attribute it to you specifically.
Said first, deliberately
The promise we will not make, and what we do instead
This belongs at the top of the page rather than the bottom, because everything after it should be read in the knowledge that the outcome is not ours to promise.
What we will do
- Find something you genuinely know first-hand that nobody in your category has published
- Publish it with a stated method, a sample description and a date on the page
- Write the central claims so each one survives being lifted out of its paragraph
- Attach every claim to a named person and to your organisation without ambiguity
- Set review dates and update visibly, recording what changed and when
- Tell you when what you hold is not interesting enough to be worth publishing
What we will never do
- Promise, guarantee or imply that any model will cite you
- Produce a statistic to be citable rather than because it is true
- Claim that markup, a file or a technical change causes citation
- Report a citation count as though it were a ranking position
- Present a single run of a prompt as evidence that anything changed
Two different outcomes
Being mentioned and being cited are not the same event
Most content aimed at assistants is built for the left-hand column and then judged on the right-hand one.
| Dimension | A mention | A citation |
|---|---|---|
| What is borrowed | Your name, as one option among several | A specific claim that you originated |
| What causes it | Being present in the sources that discuss your category | Holding something first-hand that the answer needs in order to be complete |
| Whether a link is likely | Often not. A recommendation rarely needs a source | More often, because a specific figure or definition invites one |
| What it is worth commercially | Consideration at the moment a shortlist forms | Standing as the reference point the category argues with |
| What the work looks like | Coverage, presence, consistent facts, category visibility | Original material, liftable claims, dated maintenance, clear authorship |
| How it fails | You are absent, or described from stale sources | You are quoted without attribution because the claim was paraphrased |
The work
How a business becomes a primary source
Find what you already know first-hand
Go through what your business observes and nobody records: job durations, failure rates, what customers ask before buying, what a category of work really costs this year in your market. You are looking for one thing that is true, specific, and currently unpublished anywhere.
You get: A shortlist of candidate first-hand claims
Establish it properly rather than asserting it
Decide how you know it, how many cases sit behind it, over what period, and what would make it wrong. Write that down as a method before you write the finding. A number with a method attached is a source; the same number without one is an opinion with a decimal point.
You get: A written method and sample description
Write the claim so it can be lifted
One sentence carrying the subject, the figure, the unit, the period and the origin. Then the context around it. Test it by copying that sentence somewhere else and reading it cold: if it needs the paragraph to make sense, it will be paraphrased instead of quoted.
You get: A standalone claim sentence for each finding
Attach it to a person and an organisation
A named author who genuinely did the work, with credentials that are real, on a page whose ownership is unambiguous. Then make sure the same claim is not floating in three syndicated variants with different framings, because that removes the very attribution you are trying to establish.
You get: One canonical, attributed home for each claim
Date it and commit to maintaining it
Publication date, review date, and a visible record of what changed at each update. Then only publish what you are willing to keep current. A source that quietly goes stale is worse for you than one that never existed, because it will still be drawn on.
You get: A review schedule with named owners
Give people outside the business a reason to use it
Send it to the journalists, trade publications and commentators who cover your category. Human citation and machine citation draw on overlapping ground, and a claim that no person has ever referenced is unlikely to be treated as a reference by anything else.
You get: A distribution list and a record of who used it
Where citations actually appear
The kinds of question that produce a source, not just an answer
Not every prompt invites a citation. These shapes do, because the answer is weaker without one — which is where original material has somewhere to land.
These are illustrative shapes rather than claims about frequency. Substitute your own category and run them yourself; what you are watching for is whether a source is shown at all, and whose it is.
Questions that need a number
The strongest opportunity. If no good public figure exists, the first credible one published with a method tends to become the reference.
- how much does [service] cost in [market] this year
- what percentage of [category] projects run over schedule
- average time to complete a [process]
- how many businesses in [sector] still use [older approach]
Questions that need a definition
Whoever wrote the clearest, most quotable definition tends to supply the framing the whole category then argues inside.
- what is the difference between [term a] and [term b]
- what does [industry term] actually mean
- is [common practice] the same as [related practice]
Questions about method
Practitioner knowledge, rarely written down anywhere, and the easiest category for a real operator to own outright.
- how do you decide whether [decision] is worth it
- what is the correct order to do [multi-step process]
- how do professionals check [specific thing]
Questions about what changed
Currency decides these. A dated, maintained page beats a better-written one that stopped being updated two years ago.
- what changed in [regulation or platform] this year
- is [older advice] still true
- what are the current requirements for [thing]
These are examples of how customers in this market search, drawn from keyword research and from the questions that come up on sales calls. They are illustrative, not a volume claim — the actual demand in your area is something we size before recommending anything.
Questions
What people ask before commissioning source-building work
Can you guarantee we will be cited?
No, and nobody can. Citation is decided by systems whose retrieval behaviour changes without notice, whose outputs differ between runs of the same prompt, and which no agency has any relationship with that would produce preference.
Anyone offering a guarantee here is either selling something they do not understand or counting on you never checking. What can be done is to make you the kind of source these systems have a reason to draw on. That is worth doing on its own terms, because it is the same thing that makes journalists, analysts and buyers cite you.
We have no data of our own. What can we be the source of?
More than you think. Anything you observe repeatedly and nobody else records counts: the questions your customers ask most often, how long a job actually takes, what the common failure modes are, what a category of work really costs in your market this year.
You do not need a research department. You need to write down something you already know, with a stated method and a date, in a form that can be quoted. Most businesses hold at least one such thing and have never published it.
What makes a claim easy to cite?
It stands alone. A citable claim states the subject, the number, the unit, the period it covers and where it came from, inside a single sentence a reader could lift without needing the surrounding paragraph. It also has a date on it and a named person behind it.
The test is mechanical: copy one sentence out of your best page and read it cold. If it does not survive on its own, it will be paraphrased rather than quoted, and a paraphrase generally arrives without the attribution.
Does structured data or markup cause citations?
No. Structured data helps machines interpret what a page contains and is worth implementing carefully, but there is no markup that makes a model cite you and treating one as a mechanism is how this category earns its reputation for nonsense.
Markup improves comprehension of something that already exists. If the underlying page has no original claim in it, describing it more precisely does not create one.
How would we know whether any of this worked?
Imperfectly, and never as a position. Keep a fixed set of prompts that your buyers plausibly ask, run them on a schedule, and record whether your material appears as a source rather than whether your name appears in the text. Those are different observations and only one of them is what this page is about.
Watch the non-assistant signals too, because they move first and are cleaner: whether journalists start citing the figure, whether other sites reference it, whether it gets quoted back to you in sales calls. A claim that nobody human cites is unlikely to be cited by anything else.
Is old content a problem here?
Undated content is a problem, and stale content presented as current is a bigger one. A claim about a market without a period attached cannot be evaluated, and a source that has quietly gone out of date damages you more than never having published it.
Set a review date for anything you want cited, update it visibly with what changed, and retire what you are not willing to maintain. Currency is one of the few things in this area that is entirely within your control.
Find out what you already know that nobody has published
Bring us the things your team observes and never writes down. We will tell you which of them could stand as a first-hand claim, what method it would need, and how to write it so it survives being lifted. We will not tell you it will be cited.
If we don't deliver the work we agreed to deliver for reasons within our control, you don't pay for the undelivered work. Read our guarantee
Related
Where to go next
- work aimed specifically at generated answersWhere the source-making described here becomes a programme.
- becoming the origin of a numberThe only reliable way to stop being a route to somebody else.
- binding a claim to the organisation that made it
- writing so a single claim survives being lifted
- absence and inaccuracy, diagnosed separately
- what generative engine optimisation actually covers
Last updated · Reviewed by Zubair Afzal