Can you pay to get cited by AI search engines?

Paying to get cited by AI search engines is possible in two senses that are worth keeping apart, because one is ordinary marketing and the other is a policy problem with your name on it.

The first sense is paying a person or a firm to do work on assets you own. Someone rewrites your pages so they answer the questions your buyers ask, fixes whatever is stopping a crawler reading them, and gets your facts straight on the third-party records buyers check. Nothing about that is unusual. You are buying labour and judgement, and the citation is a downstream effect you are hoping for rather than a line item.

The second sense is paying for the citation itself, as a deliverable. Vendors sell this under names like citation placement, guaranteed AI citations, and pay per citation. The mechanism is almost always publishing content about you on third-party sites that AI engines retrieve from, sometimes dozens of them, sometimes written to look independent.

No engine offers the second thing as a product. OpenAI sells advertising inside ChatGPT and documents a dedicated crawler for it, OAI-AdsBot, which visits only pages submitted as ads and checks them for policy compliance. The crawler that decides what shows up in ChatGPT's search features is a different agent, OAI-SearchBot, and the documentation keeps the two jobs separate. So anyone selling you an organic citation is selling you their ability to influence a retrieval set, not access to a product the engine sells.

Both senses cost money and only one of them is what the phrase usually means in a sales call. The rest of this page is about telling them apart before you sign something.

What are you actually buying when a vendor sells guaranteed AI citations?

A guaranteed AI citation is almost always a guarantee about a measurement rather than about your visibility, and reading the measurement is how you price the offer honestly.

Three structures turn up. A deliverables guarantee promises a quantity of published work, which is a promise about effort and says nothing about whether an engine ever reads it. A prompt-set guarantee promises a citation on specific prompts, usually chosen by the vendor, which is a promise you can satisfy by picking prompts nobody asks. A pay-per-citation model charges only when a citation is detected, which sounds like the buyer's dream and makes the definition of a detected citation the whole contract.

Each of those can be honest. What makes them hard to compare is that the engines move underneath them. Jiang and colleagues studied production ChatGPT web search across ChatGPT 5.2 to 5.5 and found that Google Search and OpenAI's own Responses API, combined, recover only about one quarter of the URLs ChatGPT cites (arXiv preprint 2610.15382, submitted 11 October 2026). A vendor verifying your citations through an external proxy is therefore measuring a minority of the thing it is being paid to deliver, whatever its dashboard says.

Four questions price the offer. Which prompts does the guarantee cover, and who chose them. How is a citation verified, and in which engine. Where will the content be published, by name. And what happens to the citations if you stop paying, which tells you whether you bought an asset or a subscription to somebody else's publishing schedule.

If the answers are vague on the third question in particular, you are probably looking at the seeded version, which the next section is about.

Does paying to seed content in AI answers actually work?

Seeding content to get into AI answers works, in the narrow sense that it produces citations quickly, and the first careful measurement of it is also the strongest argument against building a programme on it.

Liu, Hong, Zhang, Chen, Li and Yang mapped 17,211 citation instances across 6,356 unique source domains on ten AI search platforms, then tested the 22 publication platforms those cited domains sit on (arXiv preprint 2610.11932, submitted 8 October 2026). Fifteen of the 22 had low or medium barriers for both creating an account and posting. They then published posts carrying designed markers. A fabricated concept was cited by eight of the ten platforms within seven days, and the $14 package they bought was cited with its markers intact within an hour.

Two things follow, and the second is the one that should decide the purchase. The first is that citation is genuinely cheap to manufacture, so a rising citation count is not by itself evidence that anything good happened. The second is that the same paper found the top 20 domains took between 20.5% and 70.8% of citations on a given platform, and that one article on a high-preference platform carried more citation impact than more than 20 matched posts on low-preference ones. Volume on weak sites is not a substitute for presence on the handful of sites an engine already trusts.

Then there is the gap between being cited and being recommended. Lily Ray checked 100 business software queries in Google AI Overviews at three dates from April to June 2026 and found that where a brand's own best-of listicle was cited, the brand was left out of the recommendation in 224 of 323 cases, 69%. An engine can read your seeded page, use it, and name your competitor. That is the outcome the invoice does not cover.

What do the AI companies say about paying for citations?

The AI companies have started writing the practice into policy, and the dates matter if you are signing something this quarter.

Anthropic's usage policy, effective 12 November 2026, prohibits efforts to "Manipulate the sources from which search engines or AI systems draw answers" by seeding them with content that "misrepresents its origin, authorship, or independence". The policy names "networks of sites posing as unaffiliated sources corroborating the same claims" as an example, which is a fair description of how several citation-placement packages are built.

Google reached the same place from its spam policies. The Search spam policies list "attempting to manipulate generative AI responses in Google Search" as spam, alongside manipulating ranking, and define scaled content abuse as generating many pages "for the primary purpose of manipulating search rankings". A programme that mass-publishes near-identical posts about you across low-barrier platforms is the thing being described.

Note what neither policy says. Neither bans paying someone to improve your own pages, publishing a genuine guest article under a real byline, or appearing in a roundup that discloses how it was compiled. The line both draw is about concealment: whether a reader and an engine can tell where the content came from and who paid for it.

The practical reading for a buyer is that the risk is no longer hypothetical or reputational only. A tactic a vendor sold you in September can be a policy violation in November, and the vendor will not be the party whose brand is attached to the network of pages. Ask who holds that risk in writing before the engagement starts.

What can you legitimately pay for instead?

Legitimately paying for AI search visibility means buying one of three things instead of a placement, and the three are worth separating because they carry different costs, different timelines and very different ceilings.

The first is work on pages you own, which is the cheapest and the most controllable. Discovered Labs, analysing about 2 million prompt, engine and citation observations over six months across ChatGPT, Claude, Google AI and Gemini, found prompt-content alignment to be the strongest page-level predictor at a standardised beta of +0.37, 95% confidence interval +0.33 to +0.41, surviving at least four of nine robustness checks. Discovered Labs sells AEO services, so the choice of what to measure is not disinterested, and the method is unusually careful for this field. Both halves are worth saying.

The second is work off your site that you would be proud to have attributed to you: real research somebody wants to cite, a genuine byline on a publication your buyers read, accurate profiles on the review sites and directories they check. More expensive, slower, and the only version of off-site presence that survives a policy change.

The third is measurement, bought knowing what it can and cannot see. Our own category study of pharmaceutical process skid suppliers ran buying questions across ChatGPT, Google AI, Perplexity and Claude in eight categories and found no supplier won all four engines, which is the sort of thing you learn by looking rather than by reasoning about it.

How much any of it costs depends on how many pages are in scope, how much of the writing you keep in house, and whether anyone has to fix the underlying site before the content work can start. Our pricing page carries the current figures rather than this article, because an article goes stale and a quote should not.

How do you tell a real engagement from a citation-placement scheme?

Telling a real engagement from a citation-placement scheme comes down to where the work lands and whose name is on it. Six questions separate them reliably.

Question	Work on assets you own	Citation placement
Where does the published work live?	Your domain, plus named publications with real editors	Sites you have never heard of, often unnamed until after signing
Whose byline is on it?	A named person who can be contacted	Nobody, a pseudonym, or a generic staff account
Is the commercial relationship disclosed?	Yes, on the page	No, and concealment is the mechanism
What survives if you stop paying?	Your pages, your research, your corrected records	Nothing you control
What is the guarantee written against?	Scope and delivery dates	A citation count on prompts the vendor chose
Does it survive the November 2026 policy?	Yes	Prohibited where origin or independence is misrepresented

One ambiguous case deserves naming, because honest vendors sit in it. Paid placement in a genuine roundup, disclosed as paid, is advertising rather than seeding, and engines retrieve roundups heavily. Judge those on whether the disclosure is on the page and whether the list would be useful to a reader who was not paying anyone.

If you are weighing an offer right now and cannot tell which column it belongs in, a free visibility assessment will at least tell you what the engines currently say about you, which is the baseline any vendor's reporting should be checked against.

What else do people ask about paying to get cited by AI?

People ask these five questions about paying to get cited by AI once a vendor has quoted for it and somebody has to decide whether the offer is real.