In the Reuters Institute's 2026 Digital News Report, 10% of people (and 16% among under-35s) said they use AI chatbots for news each week, up from 7% a year earlier. The feature they use most is asking follow-up questions, and part of the appeal is that an answer engine can compile stories from different outlets into a single response.

So a story about your company increasingly reaches people as an AI summary, built from several outlets' coverage and whatever else the engine has read about you. The same goes for everyone else who asks about you, from buyers comparing vendors to candidates and investors doing their homework.

That adds a new job to comms. You still pitch, respond, and track coverage, but you also need to know what that combined picture says, which pages it came from, and how to change it. Nobody has time to ask nine AI engines the same twenty questions every week and compare the answers by hand.

The best AEO platforms now track that picture for you, and have added agents that take on the manual work of acting on it. Profound tracks how the main AI engines describe your brand, including the sentiment of what they say and the citations behind it, and runs its Agents on that data along with more than 2 billion real prompts. You can build your own agents in a drag-and-drop builder or start from Profound's library of agent templates.

In this article, we walk through the comms and PR jobs Profound can handle on your behalf.

Decide which outlets are worth your time

Media lists are usually ranked by reach, authority, and the relationships you've built over the years. Those are still relevant for the people who read the coverage. AI search adds another factor, the articles engines draw on when someone asks about your category, and that often reorders the list. A trade publication with modest traffic can show up in answer after answer, while a big outlet that covered you last year may never earn citations.

Three agents show you which outlets the engines lean on in your category, and which of them are worth pitching:

  • The Citation Type Breakdown for a Given Topic agent sorts the pages the engines cite by content type and reports citation counts and share of voice for each type. That tells you whether answers in your category lean on publications, review sites, forums, or vendor pages, which decides how much of your effort belongs in earned media at all.
  • The Create Media Target List agent pulls the earned media domains and pages cited in your category, with share of voice and citation counts for each, and analyzes which outlets are worth going after.
  • The Citation-Friendly PR Target Finder agent checks each cited domain's robots.txt and ranks the outlets into a pitch list, with high-citation outlets that let AI crawlers in at the top and outlets that block them at the bottom.

You can also look up any outlet's role yourself in Answer Engine Insights, which shows which websites drive AI's answers about you and how that changes over time.

Alchemy, a developer platform for crypto, uses Profound to see which trade press, Wikipedia, and reference pages shape AI answers on the prompts it wants to own, and to decide which publications are worth pursuing. As Max Crawford, Growth Marketing Manager, explains, "the sources that influence AI answers and the ones that drive traffic are rarely the same."

Find out whether a bad story is your problem or the category's

When an unflattering pattern shows up in AI answers, the instinct is to treat it as yours to fix. Sometimes it is. Others, though, the engines say something similar about every vendor in your space, because they're repeating a narrative about the category, like "tools in this space take months to implement."

Correcting that one page at a time won't change much, because the claim isn't really about you. Changing it takes a longer campaign, and sometimes it means deciding to leave it alone.

Comparing first saves you from spending a week on the wrong response:

  • Profound's Sentiment pulls out the claims engines make about you, groups them into themes, ties each theme to the citations behind it, and shows how every theme performs for you and your competitors in one view. A theme that appears at similar strength for every competitor belongs to the category. If a theme is yours alone, the citations behind it show you where it started.
  • In Chapter three of our AEO Guide, we describe market perception prompts, which ask how AI frames the category itself. These include questions like "What is a digital asset management platform and how should I think about evaluating one?" Tracking a few of them shows you the category narrative directly, before it shows up in answers about you.
  • The Top Sentiment Themes Report agent scores the themes in how engines describe you on a topic and reports your strongest and weakest themes over a period you choose, so you can tell a new problem from one that's been there all along.

Answer Engine Insights also shows the topics and prompts where competitors earn more visibility and citations than you, which helps when a bad week for you turns out to be a good week for someone else.

Respond when AI repeats something wrong or damaging

Some stories need a response, and the sooner the better. An engine that picks up a wrong claim from one article can repeat it in every answer that draws on that article, to anyone who asks, until the source changes or something more credible outweighs it.

Profound's Slack integration can alert you to time-sensitive changes in visibility or citations, so a new source shaping answers about you doesn't wait for the monthly report. Microsoft Teams works for alerts too.

FactCheck extracts the claims engines make about your brand, checks each one against the facts you've given Profound, and traces every wrong claim back to the page it came from. Four agents act on what it finds:

Every one of these is a draft that waits for your approval. In comms, that matters more than anywhere else, because a correction request in the wrong tone can turn into its own story.

Build the media list and draft the pitches

A media push can take a week of research and writing before the first pitch goes out. Agents save the most time on the research, because the targets come straight from the pages the engines already cite:

  • The Citation to Outreach Target Finder agent takes the most-cited URLs for a topic, finds each author's name, role, and contact details, and writes one row per target to a spreadsheet.
  • The Draft Outreach Email agent scrapes a target page and drafts a pitch built around what it says and what it leaves out, like a roundup that lists five of your competitors and not you.
  • The Reddit Insights Generator agent analyzes the most-cited Reddit discussions on a topic and reports what people are saying, so you know which threads are worth joining.

Profound also connects to services that handle placements. With the Noble integration, an agent finds the cited pages where you're not mentioned and passes those domains into a Noble campaign. Noble then handles outreach, content drafting, negotiation, and payment, and you review and accept the offers in your Noble dashboard.

The PartnerStack integration adds agent steps that check PartnerStack's network of more than 65,000 publishers against your citation gaps and recruit partners to mention you. And if G2 matters in your category, Profound's data powers G2's AI Visibility Dashboard, which shows how your G2 pages are cited across ChatGPT, Gemini, Claude, and Perplexity. That dashboard is in private beta for selected G2 customers.

One Identity, an identity security company, used Profound to find the publisher pages the engines cited most in its category. The team contacted those publishers with deeper product context and, where it made sense, paid for sponsored placements. One Identity won first or second position on the most influential listicles in its space, including PeerSpot, CyberPress, and Expert Insights. Its visibility grew 30% in a quarter, and its non-branded citation share doubled.

See what changed after the story ran

Coverage reports tell you where a story ran and how many people might have seen it. If you want to know whether it changed what AI says about you:

  • In Answer Engine Insights, set a date range starting from publication and compare it with the period before. Look at your visibility on the prompts the story touched, the sentiment themes attached to you, and whether the engines cite the new coverage.
  • Run the Top Sentiment Themes Report for the weeks before and after, and compare the two.

You can also build your own agent for this. Have it pull the sentiment themes on the story's topic for the weeks before publication and the weeks since, then list the new themes, the outlets whose articles the engines cite for each, and how long each took to show up. That tells you which outlets carried the story into AI answers, which is the best guide to where to pitch next time.

Allow for normal movement while you read the results. When Profound tracked 15,155 brand setups daily on Google's AI surfaces, visibility on a single engine drifted 2 to 3 points a day on its own. A change that holds for a couple of weeks is the one to report.

Run comms in AI search with your AI Marketer in Profound

Stories land on any day of the week, and no comms team can keep all engines' answers in view by hand. In Profound, that job falls to AI Marketer (Aim).

Aim is the orchestrator inside Profound, a marketing teammate that reads your AI search data, works out what's worth doing, and builds the agents to do it. Aim draws on everything Profound tracks about your brand, including visibility, citations, sentiment, FactCheck, and benchmarks against competitors. It also draws on Context Manager, which holds your company's positioning, audiences, competitors, offerings, and approved claims, filled from Slack, email, and Gong, and from the questions it asks when something's missing.

Here's how it works for comms:

  • Every week, Aim brings you a ranked list of opportunities, with the data behind each one. For comms, that might be a sentiment drop on a key topic, a falling benchmark against a competitor, or a wrong claim that's started showing up in answers.
  • You decide which opportunities to act on, and Aim turns each one into a project. Say a negative theme starts spreading from a single review. The project might include a correction request to the publisher, an update to your own page, a reply to the Reddit thread repeating it, and pitches to the two outlets the engines cite most on the topic. Aim then builds and runs the agents that execute each task.
  • Nothing goes out without your approval. Afterward, Aim tracks whether the engines' description of you changed.

Aim applies your terminology, approved claims, tone, and hard rules by default. For comms, that means a correction request drafted on a bad morning still sounds like you and doesn't make claims you haven't signed off on.

Book a demo to learn more.

AI agents for communications and PR FAQs

How do AI engines decide what to say about our company?

From the pages they find when someone asks, combined with what they learned in training. Web-searching engines pull a set of pages for each question and build the answer from them, and they tend to hold back on a claim that only one source makes.

For comms, that means a single article rarely changes the answer by itself. A message that shows up consistently across several outlets, your own site, and the review sites in your category is much more likely to end up in what the engines say. Which outlets count depends on your category and the engine, so check the citations before you plan.

Does press coverage still matter if people read AI summaries instead?

Yes, and in some ways it matters more. AI engines build their summaries from coverage, so the outlets they cite become the raw material people read about you, whether or not anyone clicks through. Coverage still reaches its own readers directly, too.

What changes is which coverage counts. An article in an outlet the engines never cite reaches its readers and stops there. The same article in an outlet they cite often can keep shaping answers long after it runs. That's why it's worth checking citations before you rank your media list.

What should we do when an AI answer repeats something false about us?

Start with the source. You can't edit an AI answer directly, but you can change the pages it draws from. Use FactCheck to find the page the wrong claim came from, then correct your own pages, ask the publisher of the third-party page for a correction, and reply in any Reddit thread repeating it. If you don't have a page that states the correct fact clearly, publish one, so the engines have something to cite in its place. Then keep tracking the prompts where the claim appeared until the answers change, since an engine can keep repeating a claim for a while after the source is fixed.

How do we measure PR's impact on AI answers?

Compare before and after on the prompts the coverage touched. Look at your visibility, the sentiment themes attached to you, whether the engines cite the new coverage, and your share of voice against competitors over the same window.