How to Get Cited by ChatGPT & Perplexity: 2026 GEO

Most of Your AI Search Strategy Lives on Someone Else’s Website

Here is the stat that should reorganize how you think about getting found in 2026. 82% of the citations AI engines hand out come from earned media. 94% come from sources you did not pay for. That is from Muck Rack’s December 2025 analysis of more than a million AI citations. Their larger May 2026 follow-up, across 25 million links from ChatGPT, Claude, and Gemini, landed in the same place at 84% earned media. Your own website? McKinsey’s August 2025 AI Discovery survey puts it at roughly 5 to 10% of the sources these models pull from.

Read that again. The page you spent three weeks polishing is a rounding error to ChatGPT.

So if you want to know how to get your business cited in ChatGPT, the answer starts somewhere uncomfortable: not on your site. AI search optimization in 2026 is mostly about what other trusted sources say about you. That is public relations. That is the whole game now.

If you already read Why Your Bakersfield Business Doesn’t Show Up in ChatGPT, you know the level-1 moves: answer-first paragraphs, FAQ schema, clean structure. Those are real. They are also table stakes. This is the level-2 post. The part nobody wants to hear because it cannot be fixed with a plugin.

Earned Media Beats Owned Content. It Is Not Close.

The numbers keep landing in the same place from people who do not work together.

• Stacker and Scrunch ran a citation-lift study across 944 prompt-and-platform combinations in December 2025. Content that lived only on a brand’s own site cited at 8%. The same content distributed through third-party news outlets cited at 34%. That is a 325% jump. A March 2026 follow-up clocked a 239% median lift across a broader sample. Same direction, different month.
• Ahrefs studied 75,000 brands in 2025 and 2026. Brand web mentions correlated with AI visibility at 0.664. Backlinks sat at 0.218. Mentions are roughly three times more predictive than links. The old SEO scoreboard is not the AI scoreboard.
• Fullintel and the University of Connecticut, presented February 2026: 47% of AI citations came from journalistic sources, and 89%-plus traced to earned media, 95% to unpaid sources.
• Moz looked at about 40,000 queries and found 88% of Google AI Mode citations were NOT in the organic top-10 for that query. AI search is a different animal than blue-link search.

Want the most independent anchors, the ones with no agency selling behind them? The academic work. Chen et al. (arXiv, September 2025) documented “a systematic and overwhelming bias towards earned media.” The GEO-16 framework from Kumar et al. (arXiv, September 2025) found that even high-quality pages “may not be cited if they reside solely on vendor blogs.” Translation: your blog can be flawless and still get skipped because it is yours.

Honest note, because we do not do brochure tone here. Most of these studies come from vendors and agencies with a stake in the answer. Treat any single decimal as directional, not lab-grade. But when Muck Rack, McKinsey, Ahrefs, Stacker, Fullintel, and two arXiv papers all point the same way, the theme is not in question. Earned media is the lever.

The Real Mechanism: Consensus Across Independent Sources

Here is why earned media wins, and it is the part that makes this a deeper playbook instead of a stat dump.

AI models do not trust a claim because you made it. They trust it because multiple unrelated sources agree. A model treats one mention as noise. It treats five aligned mentions across five different domains as a verified fact, and then it recommends you with confidence.

Think about how it reads the room. A trade-pub review says you are good at X. A YouTube test says the same. A Reddit thread backs it up. A G2 listing confirms it. A “best of” article in your city ranks you. None of those sources know each other. That independence is the signal. The model sees consensus and stops hedging.

This flips the work. It is not “publish more on my site.” It is “get the same true story told about me in enough credible places that the machine treats it as settled.” Volume matters, but consistency matters just as much. Same positioning everywhere. If your homepage, your G2 profile, a podcast intro, and a local press mention all describe you three different ways, you have handed the model three half-confident guesses instead of one verified answer.

This is exactly where public relations stopped being a nice-to-have and became the engine of AI visibility.

Where AI Actually Pulls From

If earned media is the lever, you need to know which platforms the engines actually read. The data here is solid.

• Reddit is the single most-cited domain across every major AI engine. Peec AI’s 30-million-source study (March 2026) and the 5WPR Citation Source Index (May 2026, synthesizing more than 680 million tracked citations) both put it at the top. On Perplexity specifically, the 5WPR index puts Reddit at roughly 46.7% of the top-source share, the heaviest concentration of any domain on any platform.
• YouTube is the other heavyweight. It ranks in the top tier of cited domains alongside Reddit, and on some response sets it overtakes Reddit as the most-cited social platform. Treat them as co-leaders, not a ranked list.
• After those: LinkedIn, Wikipedia, Forbes. For “best [product]” and comparison queries, G2, Capterra, and Yelp dominate.
• Each engine leans differently. ChatGPT leans heavily on Wikipedia, Reddit, and Forbes. Perplexity leans on Reddit, LinkedIn, and G2 for B2B. Google AI surfaces lean on Facebook and Yelp.

The takeaway for a real business: the conversations and listings that decide your AI visibility are happening on platforms you may have written off as “not where my customers are.” Wrong question. The question is where the model reads.

The Own Goal: You Might Be Blocking the Robots Yourself

Before any offense matters, check that you are not invisible on purpose. This is the most common, most avoidable mistake, and the fix is free.

Somebody, at some point, added a line to your robots.txt to stop AI from “stealing our content.” Reasonable instinct in 2023. In 2026 it means ChatGPT cannot see you. Roughly 25% of the top 1,000 websites now block GPTBot, up from around 5% when the crawler launched in 2023. Some on purpose. Plenty by accident.

Here is the detail almost nobody knows. OpenAI runs two separate crawlers. GPTBot is the training and bulk crawler. OAI-SearchBot is the one that powers ChatGPT’s search feature, and OpenAI exposes them as independent robots.txt controls. Block OAI-SearchBot while allowing GPTBot and you get the worst outcome possible: you help train the model but never appear in its search results. Most people do not know the second bot exists.

The crawlers to explicitly allow:

• GPTBot and OAI-SearchBot (OpenAI, ChatGPT)
• PerplexityBot (Perplexity)
• ClaudeBot and anthropic-ai (Claude)
• Applebot (Apple, Siri)
• Google-Extended (controls Gemini and AI training use)

Now the hidden killer. CDN-level blocking sits above robots.txt. Cloudflare and similar services have a bot-management layer that can override your file entirely, and Cloudflare now ships an AI-bot block that is on by default on every plan. That block runs upstream, at the infrastructure level, before a crawler ever reaches your server to read your robots.txt. So if that toggle is on, your perfect robots.txt does nothing. Plenty of B2B and ecommerce sites are blocking major AI crawlers at this layer without realizing it.

Pull up yoursite.com/robots.txt this week. Confirm none of those user-agents are disallowed. Then open your CDN or WAF settings and check the AI-bot toggle. Pure downside removal. The cheapest win in this entire post.

The Playbook: What to Actually Do

Five moves, in order of return.

1. Audit your robots.txt and CDN this week. Covered above. Do it first. If you are blocked, nothing else you do can work.

2. Run a real PR play, not just a blog. Get named in third-party publications, trade media, and legitimate “best of” lists. For a small business that means local and industry press mentions, podcast guesting, expert quotes in journalist roundups, and getting added to relevant “Top X in Bakersfield” or “Top X [category]” lists. Remember the Stacker number: third-party distribution roughly 4x’d citation rate. This is the work that moves the needle, and the work most owners skip because it is harder than installing a plugin.

3. Get onto the platforms AI reads. Authentic participation in the subreddits where your industry actually talks. A basic YouTube presence with explainers or how-tos. Verified, complete listings on G2, Capterra, and Yelp. Reddit is the most-cited domain across engines and review sites own the comparison queries. You do not need to be everywhere. You need to be on the few platforms the model trusts.

4. The listicle tactic, with a hard caveat. Pitch yourself into existing “best of” articles with a copy-paste-ready 40-to-60-word neutral blurb plus one clear differentiator, so an editor can add you in under two minutes. It works. The caveat: starting in late January 2026, Google began hammering self-promotional and AI-generated listicle farms, with documented sites losing 40% to 95% of their organic traffic through the spring and the March core update naming scaled content abuse outright. Target real editorial lists run by real publications. Do not spin up junk ones. That shortcut now backfires.

5. Do the level-1 basics so you are extractable once cited. This still matters, it is just second. Lead each section with a self-contained 40-to-60-word answer, since roughly 44% of LLM citations come from the first 30% of a page (Kevin Indig’s analysis of 1.2 million ChatGPT answers). Add FAQPage and Article JSON-LD schema, both associated with meaningfully higher citation likelihood. Drop a verifiable statistic every 150 to 200 words. Use tables and other structured formats, which get cited more often than plain prose. Keep your dateModified current. This is the website design and SEO foundation. It makes you quotable. The PR makes you cited. You need both.

If running this audit yourself sounds like a part-time job, that is because it is. The robots and CDN check, the mention-consistency sweep, the platform gaps, the schema pass. A good AI automation setup handles the monitoring so you spend your time on the earned media, not the dashboards.

FAQ

How do I get my business cited in ChatGPT?

Stop treating your website as the answer. The data is blunt: 82% of AI citations come from earned media and only 5 to 10% from your own site. Get named consistently across trusted third parties (press, review platforms, Reddit, YouTube, legitimate “best of” lists), make sure those sources tell the same story about you, and confirm you are not blocking the AI crawlers in robots.txt or your CDN. Then do the on-page basics so you are easy to quote once a source points at you.

What is the difference between AI search optimization and regular SEO in 2026?

They overlap but they are not the same game. Moz found 88% of Google AI Mode citations were not even in the organic top-10 for the query. Classic SEO chases rankings and backlinks. AI search optimization chases mentions and cross-source consensus, where brand mentions are roughly three times more predictive of AI visibility than backlinks. You still want strong SEO. It is no longer the whole strategy.

Why does PR matter so much for AI visibility?

Because AI models trust agreement across independent sources, and earned media is how you manufacture that agreement honestly. When a trade publication, a podcast, a review site, and a journalist roundup all describe you the same way, the model reads it as verified fact and recommends you with confidence. One source is noise. Five aligned sources across five domains is a citation. PR is the machine that produces those sources.

Could my own settings be blocking AI engines from seeing me?

Yes, and it is shockingly common. Around 25% of top sites block GPTBot, many by accident. The two traps: blocking OAI-SearchBot (the bot that actually powers ChatGPT search) while allowing GPTBot, and a CDN like Cloudflare overriding your robots.txt with an AI-bot block, which on Cloudflare is now on by default. Check both this week.

How long until earned-media work shows up in AI answers?

It is not instant. Models need to crawl the new sources and watch the consensus build across them, which takes weeks, not days, and compounds as more aligned mentions land. The robots.txt and CDN fix can move faster, since that is removing a block rather than building authority. Treat it like the work it is. Steady, cumulative, and worth more every month it runs.

Find Out What AI Actually Sees When It Looks at You

Most owners have never checked whether they are blocking the bots, never mind whether anyone trustworthy is talking about them. We will. Run a free audit and we will show you where you stand: crawler access, mention footprint, the platforms you are missing, and the earned-media gaps keeping you out of AI answers. No pitch. No deck. No sales call. Just what is leaking and what to fix first.

The work works. Senior people, real deliverables, a plan that ties your public relations directly to whether ChatGPT names you or your competitor.

Reliable PR & Marketing is a strategy-first marketing agency in Bakersfield, California. We run integrated SEO, PR, web, and content for founder-led companies across Kern County and nationwide. Strategy first. Execution always.