Writesonic, an AI search visibility platform that says more than 10,000 brands and agencies use it, released six new and updated features on November 5 that trace what AI says about a brand back to the page it came from. The headline addition is Outreach Agent, which finds the author of a third-party page that AI cites, drafts an email asking for a correction, sends it and tracks the reply. The marketer approves every send.
The shift matters more than any single feature. Most AI visibility tools count mentions. Writesonic is betting that brands want to know what ChatGPT, Gemini and Google AI Overviews actually say, whether it is accurate, and which page taught it. The research behind the release is Writesonic’s own, so the numbers deserve a second look, and I give them one below.
Here is the part most companies miss. Watching what AI says about your brand, your company and your own name is only half the job. The other half is a plan to act on what you find. In my online reputation management work, the clients who get hurt are rarely the ones who never looked. They are the ones who looked, saw the problem, and filed the report away.
What Writesonic Released
The release names six tools. Five tell you what is happening. One helps you do something about it.
- Sentiment breaks down how AI describes a brand by platform and topic, and now traces each negative claim to the cited page and passage.
- Fact Check flags claims AI gets wrong, such as old pricing or an outdated company description, and names the page AI cited.
- Social citations lists the YouTube channels, LinkedIn posts and Reddit threads AI cites in a brand’s category, and whether each one mentions the brand.
- Citation Decay marks cited pages as Rising or Faded on each AI platform.
- Shopping reports how products appear in AI shopping answers, down to the SKU.
- Outreach Agent contacts the people behind cited pages the brand does not own.
Social citations, Citation Decay, Fact Check, Sentiment and Shopping are available now. Enterprise customers get the first three at no extra charge. Outreach Agent is available on request inside the platform. The release lists no pricing.
“A mention count tells a brand it showed up. It doesn’t say what AI believes about the brand or where it learned it,” said Samanyou Garg, founder and CEO of Writesonic.
Checking Writesonic’s Numbers
The release makes two research claims. Both come from studies on Writesonic’s blog, and both are worth reading in full before you build a strategy on them.
YouTube and Reddit Mentions
The release says a mention on YouTube or Reddit made a brand more than three times as likely to appear in an AI answer, against 1.3 times for a mention on the brand’s own site. The underlying study, published in June, covered 2,007 brands. YouTube scored a lift of 3.19, Reddit 3.17 and Gartner 3.15. Wikipedia, one of the most-cited domains, scored only 1.42.
The authors call those numbers directional. They say the data does not prove cause, and they warn that a single Reddit mention will not reliably triple anyone’s visibility. The study also does not say how many prompts it ran or which AI engines it tested. That caution belongs in the press release too, and it is not there.
Engines Disagree
The second claim holds up better. Writesonic’s citation overlap study ran 161,286 prompts across ChatGPT, Gemini, Perplexity and AI Overviews in May and June. No pair of engines shared more than 23.7 percent of cited domains. ChatGPT overlapped least with the others, at about 13 percent.
Independent data points the same way. Ahrefs asked five AI tools the same questions daily for 20 days and found that 50 to 83 percent of citations were gone by day two. Only 4 to 7 percent survived all 20 days on ChatGPT, AI Mode and AI Overviews. Brand names held steadier than URLs.
That last finding is the useful one.
The individual pages AI cites churn constantly. What AI says about you is stickier. A Faded label on a page may mean very little on any given day. A false claim about your pricing, repeated across engines for a month, means a great deal.
Monitoring Is Not a Plan
I have covered AI brand monitoring tools before. When InLinks launched Waikay, short for “What AI Knows About You,” the pitch was fact-checking what chatbots say about a business. That was early 2025. The tools have matured. The habits of the companies buying them mostly have not.
Here is the pattern I see. A company buys a monitoring tool, sets up alerts and gets a weekly report. Nobody owns the report. Nobody decides what counts as a problem. Then all it takes is one prospect asking ChatGPT about the company and hearing an old price or a merger that never closed. The sale is gone before anyone on the team knows it happened.
The same thing applies to people. I wrote last year that your boss and your neighbor are already asking AI about you, which is why auditing your own search results comes first. Executives, founders and expert witnesses all have a name that AI describes to strangers. That description is part of the brand whether you watch it or not.
Writesonic’s Outreach Agent is interesting because it pushes the work past the dashboard. It also raises a fair question. Automated emails asking writers to change their articles can backfire. A journalist or a Reddit moderator who gets a templated correction request may write a second post about the company that sent it. The approval step matters. Use it.
A Response Plan That Works
Whatever tool you use, the plan matters more than the software. Here is the order I recommend to clients.
- Define what you watch: the brand, the company name, product names, and the names of founders and executives.
- Assign an owner: one person reads the report and decides what happens next.
- Sort errors from opinions: wrong pricing, wrong dates and wrong facts get corrected. Negative opinions get a better answer, not a takedown request.
- Fix your own pages first: update pricing, About pages and product pages before you ask anyone else to change anything.
- Correct third-party pages with evidence: send one polite, specific request with a source attached, and accept that some writers will say no.
- Earn the mentions that count: show up on YouTube, Reddit and review sites honestly, as yourself, with something worth saying.
- Recheck monthly: look for patterns across engines, not single-day swings.
Monitoring tells you there is a fire. A plan decides who grabs the extinguisher. Writesonic has built a better smoke detector, and its Outreach Agent is a decent first step toward the extinguisher. The rest still depends on whether someone at your company is willing to pick it up.