Brand mentions are what the rest of the internet says about your company. Think YouTube videos, Reddit threads, Wikipedia entries, and news stories. AI assistants like ChatGPT and Perplexity read these mentions to decide if your brand is real, trusted, and worth recommending. Your website is your claim. Your brand mentions are the world's verdict.
Maybe ChatGPT keeps naming your competitor and not you. This guide explains why. It is usually not a website problem. It is a reputation problem. We will cover which off-site platforms AI listens to, how each one is weighted, and what to do first.
Why do AI assistants care what strangers say about you?
Because an AI cannot take your word for it. Anyone can write "we are the market leader" on their own homepage. So AI systems look elsewhere to confirm it. They check what independent people say about you across the web.
This is the key shift. For years, the web ran on backlinks. A backlink is one site linking to another. But AI assistants also count plain mentions of your name, even with no link attached. These are called unlinked brand mentions. They turn out to matter a lot.
The research is striking. Ahrefs found YouTube mentions to be the single strongest predictor of AI visibility. They outrank traditional backlink authority. In other words, being talked about beats being linked to. That is a real change from old-school search.
So an AI is doing what a careful buyer does. It asks around before it recommends you. If the web is quiet about your brand, the AI stays quiet too.
How does the Brand Authority score work?
It scores your off-site presence on five platform groups, then blends them into one number from 0 to 100. Brand Authority is one of the six levers in our complete guide to GEO. It makes up 20% of the total GEO score.
Here is the recipe our audit uses. The weights tell you where to spend your effort.
| Platform group | Weight | Why it carries this weight |
|---|---|---|
| YouTube | 25% | The single strongest signal AI reads |
| 25% | Heavily used by AI, blunt and unpaid | |
| Wikipedia and Wikidata | 20% | The registry that confirms you exist |
| 15% | The professional identity anchor | |
| Other platforms | 15% | Proof you exist everywhere expected |
Each input is scored from 0 to 100. The audit then sorts your result into a tier:
| Score | Tier | What it means |
|---|---|---|
| 85–100 | Dominant | AI knows and trusts your brand widely |
| 70–84 | Strong | Solid presence, a few gaps |
| 50–69 | Moderate | Real gaps holding you back |
| 30–49 | Weak | Mostly invisible off your own site |
| 0–29 | Minimal | The web barely mentions you |
One rule guides the whole plan. Fix your weakest high-weight platform first. That is usually YouTube or Reddit, since each one carries 25%. A point gained there moves your score more than a point anywhere else.
YouTube: the strongest single signal (25%)
YouTube is the platform to fix first for most brands. AI systems read video transcripts, titles, and descriptions at a huge scale. So a brand discussed on YouTube becomes a brand the AI knows.
The audit checks two things. First, your own channel: subscribers, video count, and how recently you posted. Second, third-party mentions: other people talking about you in their videos, descriptions, and comments. The top band needs an active channel with thousands of subscribers plus many third-party video mentions. No presence scores near zero.
Here is the playbook. Create a channel. Post a few clear explainer videos. Put your brand name in the titles, the descriptions, and the spoken words. Then pursue guest spots and honest reviews on other people's channels. The full deep dive is why YouTube is the strongest signal for getting cited by AI.
Reddit: blunt, unpaid opinions AI trusts (25%)
Reddit is where real people give unfiltered opinions, and AI trusts that. It is one of the most heavily used sources in AI answers. Reddit is the single most-cited source on Perplexity, especially for "what should I buy" questions.
The audit looks at how often you are mentioned, the sentiment of those mentions, and whether you show up in recommendation threads. It also checks for an official account and any community you run.
The playbook here needs care. Join three to five relevant communities and take part as a real person. Answer questions. Respond when people mention you. You can even host an AMA, which means "Ask Me Anything" session. But do it without shilling. Pushing your product where it is not wanted gets punished hard on Reddit. See how Reddit mentions decide whether AI recommends your brand.
Wikipedia and Wikidata: the entity registry (20%)
Wikipedia is how AI decides whether your brand is a real, notable entity at all. Every major AI model is trained on it. Wikipedia is the single most-cited source in ChatGPT, far ahead of Reddit. A page there is a strong vote that you matter.
The audit checks for your own article, its quality level, and any mention of your brand in other articles. But there is a firm rule. Never edit your own Wikipedia article. That is a conflict of interest, and editors will catch it. Build the press coverage first, and let an independent editor decide you are notable.
Wikidata is the easier cousin. It is a machine-readable database of plain facts, like your founding date and headquarters. It uses an ID called a Q-number. You are allowed to create a Wikidata entry even without a Wikipedia article. Start there. Read how to get a Wikipedia page for your company and how to create a Wikidata entry for your brand.
LinkedIn: the professional identity anchor (15%)
LinkedIn is where AI confirms your company facts and team credentials. It checks your company page, your follower count, how often you post, and whether your leaders share ideas there.
There is a bonus reason to get this right. Microsoft owns both LinkedIn and Copilot. So a complete LinkedIn page feeds directly into how Copilot understands your company. The fix is simple but ongoing. Keep the page complete and active. Have your leaders post regularly. More detail lives in optimizing your LinkedIn company page for AI visibility.
What about X, Facebook, Instagram, Quora, and podcasts? (15%)
These are presence checks, not growth channels. Think of them as boxes in a "does this company exist everywhere you'd expect" list. Cross-platform consistency helps AI confirm you are one real entity. A note of caution: an abandoned account scores worse than no account at all.
Here is how the long tail breaks down:
- X, Facebook, Instagram: keep them active and consistent. They confirm the entity, nothing more.
- Quora: answers there appear in AI training data and get cited by Perplexity.
- Podcasts: a growing signal. Episodes get transcribed, and that text ends up where AI reads. Publish each episode with a full written transcript.
- Press: a mention from the last few months beats one from three years ago. Run ongoing PR for a steady drip of fresh coverage. See digital PR for AI search.
Two groups have their own extra platforms. Developer brands are also judged on GitHub, Stack Overflow, and Hacker News. There, organic discussion helps and fake hype backfires. See developer platforms and AI visibility. Public companies are judged on financial platforms like Yahoo Finance and Bloomberg. Claim those profiles and keep them correct. See AI visibility for public companies.
It's not just volume — sentiment decides direction
The tone of your mentions matters as much as the count. The audit sorts your most prominent mentions into four buckets. Positive means recommendations and praise. Neutral means plain facts and questions. Negative means complaints and "avoid this" warnings. Mixed is a blend.
This is the part owners miss. AI assistants echo the sentiment of their sources. A brand discussed mostly in a negative way gets recommended against, not for. Volume alone will not save you.
So the fix is not louder marketing. Respond to complaints in public, calmly and helpfully. Fix the real problems people keep raising. Then earn fresh positive discussion to outweigh the old negativity. Over time the balance tips back in your favor.
How do you stop AI confusing you with someone else?
You give AI clear, consistent facts about your identity. AI can mix up brands with similar names, and that creates real damage. The audit checks for three common failure modes.
- Name ambiguity. Another company or person shares your name. AI then attributes their news or scandal to you. The fix: use complete Organization schema, a thorough Wikidata entry, and one consistent name everywhere. Always write "Acme Robotics", not just "Acme".
- Ticker conflict. Your stock code exists on another exchange too. AI may quote the wrong share price. Always pair the ticker with its exchange, like "TSXV:ABC", never a bare "ABC".
- Former-name fragmentation. You renamed the company, and the web remembers both names. AI may treat them as two firms. State "formerly known as X" on your About page. Update Wikipedia and Wikidata. Redirect the old domain.
This work is part of a wider topic called entity SEO. The deep dive is entity SEO: disambiguation and cross-platform consistency.
Should you compare yourself to perfection or to your rivals?
Compare yourself to your actual competitors. Perfection is the wrong yardstick. A moderate Reddit presence is strong if every rival has zero. Being decent in an industry where everyone else is invisible is a real lead.
So the audit can scan your named competitors to calibrate your score. The strategy follows from that. Invest first where rivals are weak. That is where you can become the default AI answer with the least effort.
How do you tell machines these profiles are yours?
You link every profile back to your website, and you label them in code. This is the bridge from off-site work to your own site. Without it, AI may not connect your YouTube channel and your homepage as the same brand.
Do two things. Link each profile from your website footer. Then list them all in your Organization schema's sameAs array. That sameAs property is a hidden line of code that names your official profiles. Think of these links as verified ID cards. They let AI fold every mention into one recognized brand. The how-to is the sameAs property: linking your brand entity across the web.
Brand mentions vs backlinks: what actually counts?
Both still count, but AI leans more on mentions than classic search did. A backlink is a clickable link to your site. A brand mention is any reference to your name, link or not. Traditional SEO was built on backlinks. AI visibility gives unlinked brand mentions far more weight.
| Backlinks | Brand mentions | |
|---|---|---|
| What it is | A clickable link to your site | Any reference to your name |
| Who it helps | Mostly classic search rankings | Both search and AI answers |
| AI weighting | Still useful, less decisive | A top signal AI relies on |
| How to earn it | Outreach, content, PR | Be talked about anywhere real |
The takeaway is simple. Keep earning links, but stop treating them as the only goal. Getting people to mention you, even without a link, now does heavy lifting. The full comparison is brand mentions vs backlinks: what AI assistants actually count.
Frequently asked questions
What are brand mentions in simple terms? They are any reference to your company name across the web. That includes YouTube videos, Reddit threads, news stories, and Wikipedia. They do not need to include a link. AI assistants read these to judge whether your brand is real and worth recommending.
Why does ChatGPT recommend my competitor and not me? Usually because the web says more about them. AI echoes what independent sources say. If your competitor has more YouTube mentions, more Reddit goodwill, and a Wikipedia page, the AI leans their way. The fix is off-site, not a tweak to your homepage.
Are brand mentions more important than backlinks now? For AI visibility, yes, mentions carry more weight than they used to. Backlinks still help classic search. But AI assistants count plain, unlinked mentions heavily. Ahrefs found YouTube mentions to be the strongest predictor of AI visibility, ahead of backlink authority.
Can I build brand mentions quickly? Some parts move faster than others. A Wikidata entry and active social profiles are quick. YouTube and Reddit presence take steady, honest effort. Wikipedia can take months and is never guaranteed. Treat it as raising your odds, not a switch you flip.
Do negative mentions hurt my AI visibility? Yes. AI echoes the tone of its sources. A brand discussed mostly in a negative way gets recommended against. Respond to complaints in public, fix the root problems, and earn fresh positive discussion over time.
Where to start
You cannot build every platform at once. Here is the order that moves your score fastest:
- This month: create a Wikidata entry and tidy your social profiles. Link every profile from your footer and your
sameAsarray. Fix any name or ticker confusion. - This quarter: build a YouTube channel and join the Reddit communities that matter to you. Take part as a real person, never as a salesperson.
- Ongoing: run steady PR for fresh press, pursue a Wikipedia presence, and answer complaints in public.
Remember the one rule. Start with your weakest high-weight platform, usually YouTube or Reddit. That is where points move fastest.
This category changes slowly, which is exactly why it pays to track it. Our visibility audit scans all of these platforms and sorts the sentiment of your most prominent mentions. It then tells you which single platform to fix first. Brand authority is 20% of your GEO score, and a monthly re-check shows whether your effort is landing.
The honest truth is that this is a big job. That is also why it is worth doing. Most of your rivals will not bother. Run a free visibility audit to see exactly where you stand and what to fix first.