Overview
It's definitely been a chaotic last couple of years, with ChatGPT storming onto the scene, forcing Google's hand, making them release Google Now Google Assistant Apprentice Bard SGE AI Overviews/AI Mode. (Google, do you REALLY have to rename everything about once a quarter???). As of January 2026, 37% of consumers start their search with AI instead of Google, according to this study by Search Engine Land. And for those that DO start at Google, about 25% result in an AI Overview answer at the top, according to this study.
Add in Anthropic and Claude, Perplexity, China's low-budget DeepSeek, Grok, Microsoft's Copilot, and the waters just keep getting muddier.
The good news is: despite the array of choices of AI answer engines (plus the multitude of versions and pricing for each), the core stuff you need to do to get cited by the AI engines is roughly the same. And even if it wasn't: today the market is mostly owned by Google's AI Overviews (powered by Gemini) and ChatGPT (from OpenAI).
The New Two-Step Search Path
In the olden days (2 years ago!), people would do a Google search on a non-branded keyword...Google would show them 10 websites (plus ads and a whole lot of other nonsense), and the person would click on one or more of those websites to learn more, and maybe buy something.
Today, that pattern has changed--and it's messing up everyone's analytics and attribution mechanisms. That first non-branded keyword search returns an AI Overview (or, they start their search in ChatGPT), and the AI answer engine runs that regular search (and, 10-20 other closely related searches, called "query fan-out"), and analyzes the top 10 to 20 results for each of those related searches. They then use their AI wizardry to figure out what's credible/correct, crank out an answer, and...if you're lucky...cite a couple of products or brands or relevant content pages from a few of those websites.
Those citations in the AI answer often contain links--but not always. Even when they do, they tend not to get clicked much. Instead, the user sees a product or company name that it recommended in its answer, and does a second search, this time for that name. Overall impressions counts in Search Console go up (twice as many searches as before), clicks stay the same (if you're cited) or go down (if you're not).
How to Get Cited in Google AI Overviews & ChatGPT
So that's the new goal: to get cited in the answers, which often turns into a search on your brand--usually VERY easy to rank for.
You might think you don't need to rank in Google anymore--you just need to get cited! True, but to get cited...you have to show up in the regular organic results in the first place, to "make the team". Otherwise you're not in the pool of web pages the AI engine will consider for citing in its answer.
So that means you still have to do all the usual "right stuff" for SEO. Sigh. But there's more to it than that.
Best Practices for Getting Cited by the AI Answer Engines
Crawlability and digestibility are key: like with Google, you've got to let the AI bots crawl your site. If your feelings get hurt because they're stealing and regurgitating your content, get over it...the alternative is that they steal and regurgitate (and link to, and mention the brand names of!) your competitors instead.
Digestibility here means schema markup, plus some good structuring of the content itself. Schema markup is more important to the AI engines than it has been to Google, so you'll want to carefully markup up what's on your page to have the best chance of the AI LLMs understanding the entities on the page. Those entities are things like: companies; products; articles; people (e.g. author of the article); reviews; services; etc.
A summary at the top of the page is important as well, as the AI bots, in their first pass through the 200-400 candidate pages returned by regular search, will read the first bit of the page to try and determine whether or not the page likely has the info they're looking for. You need to tell them yes, it does! in some of the first few sentences on the page. You ALSO need to tell them who wrote that content, and what their credentials are, so a short bio and links to their LinkedIn can help.
Beyond that: make smart use of the H1, H2, H3 heading hierarchies to help the bots understand what text block is an answer to what question. Use tables and lists--these are easy to digest and extract and show in the answers. Use an FAQ question and answer section.
Oh, and then show them that you're popular...
Popularity/Expertise Signals: More Than Just Links
Google's been positively giddy over links for decades, and SEOs--knowing this--have served up, bought, stole, faked, and firehosed links as a result. In the last few years, Google's sweet tooth for links has diminished somewhat, but it's still there. The AI engines consider links as credibility/popularity signals, but they're much more circumspect. They pay more attention to the credentials of the site that's linking, and overall, links appear to be a much smaller part of the equation than with Google.
Brand mentions matter, whether they're linked or not. Particularly in (fairly) trusted sites like Reddit, or industry publications, or newspapers.
Social media popularity matters. If lots of people are talking about your company and/or your product, that's a sign that you're hot (at the moment). Links from 2010 simply say you used to be hot.
But It's Not Just AI: Social Search is Huge!
Here's where Gen Z starts searches:
- Google: 67%
- YouTube: 68%
- Instagram: 65%
- TikTok: 58%
- ChatGPT: 34%
Of course, there's overlap there--any given person might search fashion or travel topics on Instagram, how-to videos on YouTube, "best ****" on ChatGPT, and look for coffeeshops nearby on Google.
And that doesn't even count people asking Alexa questions...saying "Hey, Google" to their Google Mini, or begging Siri on their iPhone to answer a simple question. Luckily, there's hope in sight in the near future, as Apple and Google have announced that Gemini will power Siri soon.
Real Marketing Matters
Social media is no longer just a place where you publish content in a desperate attempt to get people to click through to your website. Now, it's also a place to publish truly interesting and engaging content, getting people to like/comment/follow/share, growing your brand's awareness but possibly more importantly, showing the AI answer engines that you (and your content) are popular.
More than 10 years ago, Wil Reynolds coined the term RCS (real company sh**), telling the SEO industry to stop trying to game Google, and instead publish stuff that people genuinely were interested in. Today, that's even more true. You can't just game ChatGPT and AI Overviews with links and fake Instagram followers--you have to actually make people care about (and engage with!) what you have to say.
In Summary
- Search now happens all over: ChatGPT, Instagram, YouTube, TikTok, Siri, Alexa, and still Google
- Basic SEO best practices still apply
- Schema markup matters more than ever
- Structure matters more than ever
- Popularity & engagement matters
- Be real, be valuable, be loveable, be useful!
