AI Search Optimization: How Publishers Can Thrive in the Age of Generative Search

News travels fast, but it doesn’t always travel well.

Think about how quickly a juicy piece of gossip can spread via word of mouth, for example. Much like a game of broken telephone, important details can get misconstrued or left out entirely. That’s the price you pay when you don’t get the news from the source.

Media organizations have spent the last two decades adjusting to a model whereby the news they publish largely comes to audiences in a similarly indirect way. 

First came search traffic, driven predominantly by Google. More recently artificial intelligence (AI) searchbots have put the source of news at a further remove by scraping content and creating automated summaries of in-depth stories.

The impact of AI searchbots on traffic has been difficult to quantify, until recently. As more data comes to light, however, publishers need to rethink everything about how they position the stories they tell to get noticed by humans and algorithms alike.

What the numbers say about AI searchbots and traffic

In March, Forbes published an exclusive look at research from Tollbit based on an analysis of more than 160 news sites and blogs. It found AI-powered search engines were scraping these sites two million times, on average.

Despite all that activity, AI search engines are sending 96% less referral traffic to news sites and blogs than traditional Google search, according to the research.

The impact on publishers’ ability to monetize and grow subscribers is obvious, but it gets worse when you see how readily audiences are accepting AI-generated article summaries.

In its 2025 Consumer Adoption of AI Report, Askattest found Gen AI search results are deemed to be more trustworthy than paid ads by 40% of consumers. Only 17% said they trust AI results less than the links that regular search engines return. This may refer to information about products versus articles covering current events, but you can’t necessarily count on audiences to dig deeper.

When AI search tools do cite news, they’re often incorrect. In March, Columbia Journalism Review published research from the Tow Center for Digital Journalism, which conducted tests on eight generative search tools with live search features. They used a random selection of 10 articles from 20 different publishers to identify the source of a fact or a quote.

AI chatbots cite the wrong news source 60% of the time, the researchers found, and some had error rates as high as 94%. Free models performed even worse than the premium versions, the report added.

How publishers are responding to the AI searchbot threat so far

In early May, the Wall Street Journal reported that some companies believe their websites will inevitably become simply data sources for large language models (LLMs). However the same article cited data from market research firm Forrester that put the threat into perspective: Google still handles five trillion searches per year, dwarfing the one billion searches by a leading generative AI firm.

Media companies aren’t willing to wait and see how quickly those numbers change. Hundreds of publishers who make up the News/Media Alliance banded together in April to launch an ad campaign dubbed “Support responsible AI,” which is aimed to raise awareness about the way the technology threatens the industry.

Beyond demands to be compensated for the data they feed into AI searchbots and that regulars monitor for anti-competitive practices, the campaign also mandates transparency, sourcing, and attribution in AI-generated content.

Google has already begun addressing some of these concerns with AI Mode, which weaves links and citations directly into its searchbot results. An article on SEO Roundtable also quoted a Google VP who said the company will be considering additional features that provide publishers new opportunities for their content to rank.

Strategies for getting stories to show up in AI search results

Media organizations can continue to build up channels that provide them direct access to audiences, but ignoring or trying to avoid AI searchbots isn’t a viable response. Instead, it’s time to begin learning more about what is variously being called expanded search, decentralized search, and AI search optimization (AISO).

Though adapting in this environment can feel a bit like building a parachute while falling out of an airplane, these are some of the areas to explore and experiment:

1. Use keyword research to think through conversational prompts

Audiences are still bringing up news sites and scrolling through the top headlines, but publishers have long realized a portion of them will land on stories based on keywords they’ve entered into Google. These could be one or two terms that relate to the topic of an article, and we’ve all gotten used to Google figuring out what kind of content users want to see in results pages.

AI search is fundamentally different in that it can be based on a conversational interface where audiences are asking very detailed questions, in full sentences. This means getting noticed depends on having content that supports all the contextual details that get included in a prompt.

Let’s say your editorial team is planning a story package aimed at parents who want to get a head start on shopping for back-to-school items. Keywords like “back to school” or “school supplies” might have once been enough to work from a search engine optimization (SEO) perspective. If your target audience is using AI search engines, on the other hand, they might enter a prompt like this:

I am a single working mother based in Newark, NJ with three children all under the age of 13. I need to figure out everything they’re going to need to be successful in grades 3, 5, and 7, and all the supplies need to cost less than $200. Recommend the best choices for me.

Notice what this prompt includes:

• Target audience, or what other businesses would call their ideal customer profile (ICP)
• Their specific pain point (looking for school supplies for multiple children at once, all of whom might have different needs)
• Demographic information about the searcher
• Location
• Budget

Developing a story package can begin with keywords but should become much more specific, including product categories. This is not about stuffing an article with keywords but including everything that will make an AI searchbot add snippets from your story package into its answer. This could lead the person searching to dig deeper and click through to the story package on your publication’s lifestyle section.

2. Use analytics to identify what’s bringing AI searchbot traffic to your site

Even if only a small percentage of people click through from an AI-powered search result to a publisher’s site, you should aim to understand what topics drove them there.

This is another good reason to make sure you’ve got a streamlined approach to analyzing all traffic sources. The Google Search Console’s Parse.ly integration, for instance, saves you from having to log into separate applications and provides greater visibility into what’s happening from a search traffic perspective. From there, you can put more time into optimizing for AI search.

3. Assist AI with your own summaries

Editorial teams tend to be highly skilled in winnowing down stories so the most important details stand out. This is how reporters pack in the most compelling facts in their lede when they’re writing a news story, for instance. You can take a similar approach in helping both audiences and AI searchbots see at a glance the value your content is offering.

Gannett, for example, has introduced highlights at the top of each story it calls Key Points, which are instantly generated as journalists finish writing. This has already led to higher engagement on some articles, and could make Gannett’s stories stand out within AI search engines as well.

4. Demonstrate your authority

Publishers with a long history can enjoy a strong reputation among their existing audience, but AI searchbots may be trying to determine credibility on a story-by-story basis.

An article offering a product review of a new smartphone might benefit from an expanded author bio that details how long they’ve been studying the consumer electronics industry and how many other reviews they’ve written. An op-ed might need to spell out the contributor’s educational background, career history, and other proof points that signal to AI they’re legit.

This is not unlike the criteria Google’s human reviewers have been using to judge content quality. Experience, expertise, authority, and trust (EEAT) should be evident on every page. As Search Engine Journal recently reported, Google has updated its guidelines to avoid attempts to fake EEAT. Never take your reputation for granted, or it might get buried behind an AI searchbot result.

5. Publish with follow-up questions in mind

The International Journalist’s network (IJNET) has some helpful advice about humanizing your audience. Instead of thinking in terms of visitor or subscriber numbers, for example, IJNET recommends building a narrative around your target audience before the editorial team writes or develops any content.

“Who is this story for? Who should I talk to? What context do they need? What format will this work best in? When will they be most receptive? And, finally, how can I find those readers online?” the article said.

This is something journalists have always been good at doing. Whether it’s a press conference or a one-on-one, they are masters at asking the follow-up question that makes their story stand out from the competition. Now, they need to use that skill by imagining they’re a member of their audience chatting with an AI search tool.

Publishers’ stories not only have to deliver on the prompt, but whatever comes next. AI searchbots may be taking over a lot of the effort in finding answers, but media organizations will only survive by keeping the interests of the people using these tools at the center of everything they do.