The fastest way to change how AI talks about your brand usually is not on your own website. It is on the sources models already trust. A citation, where the model uses a page as its source, carries more weight than any line you write about yourself, because AI models weight independent mentions far above your own marketing copy.
Owned content is what you control. Earned citations are what models trust. You need both, and you should start by finding out which pages models cite today.
Why earned beats owned here
A homepage tagline is a claim. A balanced review on a respected platform is evidence. Models behave like a skeptical buyer: they discount what a brand says about itself and lean on what independent sources say. That is why the off-site map below usually moves the needle more than another landing page.
It also explains why you cannot copy one playbook everywhere. Similarweb reports, from roughly 600,000 AI citation events, only about 11% domain overlap between the sources ChatGPT and Perplexity cite, and about half of cited domains changing month to month (Similarweb, a vendor-published figure). The sources that matter differ per model and per month, so the target list has to come from data, not from a generic list.
The sources worth earning
Not every citation is equal, and models appear to discount obviously promotional content. Aim for credible, independent surfaces:
Source type | Examples | Why it gets cited | Typical effort |
|---|---|---|---|
Review platforms | G2, Capterra, TrustRadius | Recent, specific customer reviews, strong in software categories | Ongoing review campaign |
Editorial and press | Industry publications, trade media | Independent authority | Digital PR, expert contributions |
Communities | Reddit, Hacker News, niche forums | Buyers compare notes in plain language | Genuine participation, slow |
Reference sites | Wikipedia, Wikidata | Structured, neutral | Only if you meet the notability bar |
Comparison pages | "Best of" and "X alternatives" pages run by others | Directly answer buying prompts | Outreach with accurate facts |
Do not fake notability on Wikipedia; the criteria are strict and gaming them backfires. Social platforms can matter more than you expect on some engines: Promptwatch reports they were 0.46% of ChatGPT citations but 12.2% of Google AI Overview citations in one week in autumn 2026, as summarised by Orchly. Treat figures like this as a prompt to check your own data, not as a rule.
How to actually earn them
Treat this as a targeted campaign, not scattered outreach:
Find what AI already cites for your category. Take your 10 to 20 most important prompts and list the exact URLs cited per model. Those pages are your priority list.
Find the source gaps. Look for cited pages that mention your competitors but not you. These are the highest-value placements because the model already trusts the page.
Rank by signal and effort. A comparison list that appears in three prompts beats a forum thread that appears in one.
Earn the placement honestly. Run customer review campaigns, pursue digital PR, contribute real expertise, and ask list maintainers to include you with accurate facts. A glowing affiliate post does less than a balanced write-up from a source readers and models trust.
Fix your own pages too. Citations rarely stand alone. Clear definitions, schema markup and a crawlable site make it easier for models to reuse what others say about you. See how to automate schema markup for AI citations.
A caution on scale: not every cited page is a realistic target. Ahrefs data summarised by Orchly suggests only about 32.3% of ChatGPT's top 1,000 cited pages are realistic outreach targets. Prioritise the ones where a genuine contribution is possible.
Close the loop
Earning a citation is not the finish line. The point is whether it changes how you appear in AI answers, and the only way to know is to measure before and after. Record your baseline Visibility Score, Share of Voice and citations, earn a placement, then watch whether your presence moves over the following weeks. Because Trackbase re-runs your prompts every 24 hours across up to 11 AI models, you can see a new source appear in the cited sources view and tie it to a change in presence. Trackbase shows which sources are already feeding your category and whether your earned coverage is paying off; the cited-sources view is part of the Professional plan.
Pair this with the on-site work in the playbook for improving your AI visibility, and you cover both halves of the equation: the content you own and the trust you earn. If you are comparing tools to run this process, read how to choose among AI citation tracking alternatives.
FAQ
What does it mean to be cited by AI?
An AI answer credits a specific URL as a source. This is different from a mention, where your brand name appears in the text without a source link. The two vary independently, so track both.
How long does it take for a new placement to show up in AI answers?
There is no fixed timeline. It depends on the model, on whether it retrieves live sources, and on how often the page is crawled. Measure daily and compare against your baseline over several weeks before judging a placement.
Should I buy placements or sponsored reviews?
No. Models and readers discount obviously promotional content, and platforms penalise undisclosed paid reviews. Invest in genuine customer reviews, real expertise and accurate listings instead.




