This article is part of our guide on what makes AI recommend a brand.
Ask ChatGPT for the best options in your category. Three competitors come up. Now open the sources under the answer: a comparison on a trade blog, a ranking on a review site, an article in an industry magazine. Your brand is on none of them.
That is the whole mechanism, in one screen. The AI engine did not decide on its own that your competitors were better. It read a handful of pages and repeated what they said.
For twenty years, Google authority came from backlinks: other sites pointing to yours. In AI answers, authority comes from mentions: other sites naming you, on the pages the AI engines actually read. The principle is the same. The playbook is not.
Why do AI engines care about what other sites say?
AI engines cannot trust what a brand says about itself, for the same reason Google never could. Every homepage claims to be the best. So before answering a buying question, the engine searches the web, opens a few pages, and builds its answer from them.
Those pages are rarely brand sites. They are comparisons ("best CRM for a mid-size sales team"), rankings, review platforms, directories, trade media, sometimes forum threads. If they name your competitors and not you, the answer does the same, whatever you publish on your own site.
This is where the backlink parallel holds. In SEO, a link from a relevant, trusted site was a vote. In AI answers, a mention on a page the engine cites is a vote too, and a more direct one: the engine quotes that page while it writes the answer your buyer reads.
There is also where the parallel breaks. A backlink counted wherever Google found it. A mention only counts if the AI engine opens that page on your questions. A link on a site with a high authority score that no AI engine ever cites on your topic moves nothing.
Which sites do AI engines actually cite in your market?
Fewer than you would think, and not the ones you would guess.
On a given question, the same domains keep coming back from one day to the next: a few comparison sites, one or two review platforms, a trade magazine, sometimes a directory nobody on your team has heard of. Those domains are not the same everywhere. ChatGPT and Perplexity do not read the same sources. A question asked in France is answered from French sites, the same question asked in Germany from German ones.
That is why a generic "top 100 sites for digital PR" list does not help. What you need is the list of pages cited on your questions, in your countries, with the competitors each page names.
Mentionable records every source cited on each of your tracked questions, every day, by the AI engines you choose among ChatGPT, Claude, Perplexity, Gemini, Google AI Mode and Google AI Overview. On the sources page, you see which domains and which exact pages shape the answers in your category, by AI engine and by country. From that data, Mentionable isolates the ones worth getting: pages cited on your questions where a competitor is named and you are not.
That list changes the conversation in a marketing meeting. "We need more PR" becomes "these twelve pages are cited on our main questions, nine of them name our two main competitors, here is who to contact on each".
Mentions versus backlinks: what changes for your team?
Three things change, and they matter for how you staff it.
The target list is shorter and more precise. In link building, more links from more domains was usually better. For AI visibility, a dozen pages that the engines cite on your questions are worth more than a hundred links on sites they never open. Precision beats volume on the target side.
The context matters as much as the presence. A mention counts when the page names you for the right thing: in a comparison of your category, in a ranking for your use case, in an article about the problem you solve. A passing mention in an unrelated listicle does little.
The work is closer to digital PR than to SEO. Getting a brand added to an existing comparison means writing to an editor or an author with a reason to update the page: what your brand does differently, for whom, with what proof. That is relationship work, and it is the part most marketing teams never get to, because it takes hundreds of messages and follow-ups every month.
How do you get mentioned on the sites AI engines cite?
Most of the work fits in four moves.
Start from the pages, not the domains. Take each cited page where a competitor is named and you are not. Note which questions it is cited on and what it says about your competitors. That gives every message a precise reason to exist.
Give the editor a reason to update. Editors update comparisons when the page is getting traffic or citations and something new is worth adding. Tell them which questions their page answers in the AI engines, what your brand does that the listed options do not, and offer what they need: a demo account, figures, a short description they can paste.
Follow up, then follow up again. Replies often come after the first follow-up. Without a system, follow-ups are the first thing a busy team drops.
Keep your own site ready for the visit. An editor who checks your site before adding you needs to find a clear offer, prices or a pricing logic, and proof. The same goes for the AI engines that read your site once the mention sends them there. That side of the work is on your site: see why your website is invisible to AI.
None of this is complicated. It is just a lot of messages, every month, to the right people. That volume is what stalls most teams.
What does Mentionable's outreach do for you?
From the Influence plan, Mentionable runs the outreach campaigns to the sites the AI engines cite on your questions: 200 sites contacted per month on Influence, 500 on Reference. Here is who does what.
| Step | Mentionable | Your team |
|---|---|---|
| Pick the targets | Lists the pages cited on your questions that name a competitor and not you | Reviews the list if it wants to |
| Find the contact | Finds the editor, the author, a partnership address or a contact form | Nothing |
| Set up the sending | A dedicated domain and warmed-up mailboxes, so messages land in the inbox and your main domain stays safe | Nothing |
| Write the messages | One personal message per site, naming the cited page and why your brand belongs there, in your brand voice | Nothing |
| Send and follow up | Every month, as new sites start to be cited on your questions | Nothing |
| Handle replies | Reads and sorts every reply, answers the simple ones | Decides on what needs a decision: a price, a product question |
| Close the mention | Follows each mention until it is published | Handles the agreement |
You activate outreach from the app, and the Mentionable team contacts you to set everything up. During the 7-day trial, nothing is sent, but the Action plan already shows the sites that cite your competitors and not you: the targeted sites and the message Mentionable would send them.
When does buying a link make sense?
Buying links is not the strategy here, but it is not forbidden either. It is a tool with a narrow use.
A sponsored article or a link insertion is worth considering when the site is already cited by the AI engines on your questions and sells placements. You get onto a page the engines read in days rather than months. Mentionable crosses the sites cited on your questions with the catalogs of netlinking marketplaces, so the Backlinks page shows which cited sites sell a link, at what price, ranked by their weight in AI answers.
Buying links on sites no AI engine reads, because they have a nice authority score, does nothing for your AI visibility. And a paid placement is never as strong as an editor deciding to add you to a comparison they maintain.
Most brands end up doing both: buy where a cited site sells and the price is worth it, run outreach everywhere else.
How do you know a mention changed the answers?
This is the question your director will ask after the first quarter, and it deserves a number rather than an impression.
AI answers move on their own. The sites cited on a question often move from one month to the next without anyone doing anything. So a rise in mentions after a new article could be your work, or just the market moving.
Mentionable separates the two. When you mark a source as done in the Action plan, the Progress page measures it on the questions it targets: the share of AI answers that name your brand, the 4 weeks before against the 4 weeks after, compared with the questions you did not touch over the same dates. The verdict reads "It worked", "Being measured" or "No visible effect". A press article obtained outside the tool can be logged by hand, and it shows up on your visibility chart so you can see what happened after that date.
Mentions also compound. A brand that appears in the comparisons gets recommended more often, gets discovered by more buyers, and ends up in the next round of articles written about the category. The opposite is true as well: the longer the cited pages ignore you, the more the gap with the competitors they name widens.
Where to start this week
Pick your ten most important buying questions. Look at the pages the AI engines cite on them, country by country. Count how many name a competitor and not you. That number is your Off-site backlog, and it is usually longer than any team can work through by hand.
The 7-day trial needs no credit card. Enter your URL, keep the questions that matter, and within a few days you see which sites shape the answers in your category, and what Mentionable would send to them.
If your team would rather hand the whole thing over, Mentionable Agency goes further than the tool: its team builds the strategy, writes and gets the links and mentions with its own methods, and the tool measures your visibility in AI answers before and after.
Related articles
- What Is an AI Citation?: how AI citations differ from mentions and why both build authority.
- AI Source Authority: the signals that decide which sources AI engines trust and cite.
- GEO strategy: the 3 levers: measurement, your site and outside mentions, in one plan.
