When Cision ran more than 350 variations of the question “what is the best car” through five AI platforms, YouTube showed up as a cited source in 4,247 answers. Toyota’s own website, belonging to the brand those same answers recommended most, showed up in 135.
That gap is the most useful number in The State of Search and Social 2027, a report Cision released on November 9. It means the pages a brand controls are a small slice of what AI reads before it describes that brand to a buyer. Keep the source in mind. Cision owns the social listening company Brandwatch and sells the software that connects search, social and AI data, so the report doubles as a case for buying it. The data set is still large: 564.61 million social posts and 4,678 search queries and AI prompts.
The practical step is to find out who is actually writing your AI answers. In my online reputation management work, the first question is no longer only where a client ranks. It is who else is talking about them, on which platforms, and whether any of it is accurate. Creators, reviewers and journalists are supplying that material right now, with or without your input.
What Cision Measured
The report draws on Brandwatch Consumer Research and Trajaan data. The press release gives the research period as August 2025 to August 2026. The report’s own methodology says September 2026. The full report is gated behind a sales form.
Amy Jones, Chief Marketing Officer at Cision, summed up the premise:
“Search can show where interest is changing. Social helps explain why. And AI answers reveal the narrative reaching consumers.”
Brandwatch made a related case in June with its Question Gap research, which found social questions run about 20 times longer than search queries. This report adds the third signal: what AI tells people once they ask.
The Car Data Is the Story
Toyota led the “best car” answers with a 13.8 percent share of brand mentions, appearing in 5,957 answers. The report scores AI sentiment toward Toyota at 90 percent positive.
The sources behind those answers tell a different story. YouTube held 9.3 percent of cited domains. Then came whatcar.com, cars.usnews.com, edmunds.com and carwow.co.uk. Three motoring writers, Darren Moss, Ryan Birch and Mario Christou, appeared in almost 3,000 responses combined. No carmaker made the top ten.
Two details stand out. The press release says YouTube had “more than 4,000” sources to Toyota’s “fraction,” but the report counts answers, not sources. And four of the ten most-cited domains are British: whatcar.com, carwow.co.uk, autoexpress.co.uk and carmagazine.co.uk. The search data covers the US and UK together, and the report does not split the AI results by country. A buyer in Dallas may get a different set of sources than a buyer in Leeds.
Fewer Searches Does Not Mean Fewer Buyers
US and UK Google searches for “best car” and related terms peaked near 840,000 a month in March 2025. By August 2026 they were running close to 420,000. The report reads this as buyers taking the question to AI, and it points to new car registrations trending upward globally.
That is a reasonable read. It is also a correlation. The report shows searches falling and AI answers existing. It does not measure buyers moving from one to the other.
I have heard “our keyword traffic is down” from clients for most of my career. Sometimes demand moved. Sometimes only the question moved. A chart of falling search volume cannot tell you which one happened.
When Customers and AI Describe a Product Differently
The health section follows magnesium glycinate, which overtook melatonin in Google search volume in February 2026. Melatonin mentions spiked above 17,000 in early November 2025 after coverage of an American Heart Association study on long-term use and heart failure.
On social, stress relief accounted for 46 percent of the benefits people discussed. In 1,032 AI responses collected between August 15 and September 14, 2026, bone support led at 30 percent and cardiometabolic support followed at 28 percent. Stress relief drew 13 percent. Customers buy for one reason. AI pitches them another.
Check Which Models Were Asked
The methodology is worth reading here. Those 1,032 responses came from 130 prompts run across Gemini 3 Flash Lite, Grok 4.1 Fast, GPT-5.4 mini, Claude Haiku 4.5, Perplexity’s Sonar, Google AI Overviews and Google AI Mode. Apart from Google’s two search features, most of those are the lighter, lower-cost versions of each model family. The report does not say they match what a shopper sees in the ChatGPT or Gemini app, and it does not name the five models used for the car study.
That does not sink the finding. It does mean you should run your own prompts in the tools your customers use before you rewrite a product page.
Interest Is Not the Same as Approval
The report says searches for AI-related terms rose 2,400 percent after ChatGPT launched in November 2022. Read the chart label before you repeat that number. The query set includes “Elon Musk” alongside OpenAI, Copilot, Gemini and Claude. A celebrity name in the bucket inflates the curve.
“AI slop” drew 537,000 searches from January to August 2026, up 272 percent from the same months in 2025. About 84 percent of social mentions were negative, though the sentiment charts leave out neutral mentions. Cision also tracked 11,400 mentions of canceled subscriptions and uninstalled apps blamed on AI slop, up 1,900 percent year over year. The report says people object to having no choice and getting nothing of value, not to AI itself.
Publishers producing AI content at scale should take note. Google’s guidance, which I covered when Google confirmed that AI SEO is still SEO, favors original insight over commodity content.
The Reputation Problem Search Never Showed
Data centers are the clearest case in the report. The topic barely registers in search. Social mentions passed 400,000 in May 2026, and anger made up half of all emotion-tagged conversation. Most came from people living near construction, noise and environmental damage.
A search-only dashboard would have shown nothing. Katie Slayford, Senior Social Media Manager at Brandwatch, said teams need to monitor “their industry’s broader impact across channels, not just mentions of their brand.” I agree. The reputation crisis you see coming is rarely the one that hurts you.
Viral Versus Always-On
Dubai chocolate peaked near 8 million monthly US and UK searches in early 2025 and slid to about 1 million by August 2026. Matcha held between roughly 6 and 8.5 million the whole time. The “maxxing” trends show a third pattern, with search cooling as social keeps growing. Searches for maxxing terms fell 55.7 percent over the three months to August 2026, but looksmaxxing mentions on social rose 807 percent year over year and tanmaxxing rose 1,500 percent. Fast growth is not the same as brand fit.
What I Would Do With This
I have argued that optimizing for AI is still SEO. This report does not change my view. It sharpens it. Entities, mentions and citations decide AI visibility, and most of the citations live on sites you do not own.
When I wrote about Writesonic’s AI brand monitoring tools last week, its study of 2,007 brands found a YouTube mention carried a 3.19 lift in AI visibility. Its authors called that directional, and so should you. Two vendors with different methods still landed on YouTube.
Here is where I would start:
- Run the questions your buyers ask through the AI tools they use and record which sources get cited.
- List the creators, reviewers and journalists who appear, and treat that list as your real search results.
- Check whether those sources describe your product, pricing and strengths accurately.
- Compare how customers describe your product on social with how AI describes it.
- Treat falling keyword volume as a question to investigate, not a verdict on demand.
Slayford put the larger point well in the report: social media is “part of the raw material that shapes what AI says about your brand when a buyer asks.”
Toyota earned its 13.8 percent mostly on other people’s pages. Most brands have never checked whose pages are carrying theirs.