2015: you were a list

In 2015 a search for a company or a person returned ten blue links, and the stranger who typed the name did the reading. Your site, a LinkedIn page, and if you are lucky - whatever a newspaper had once written about you sat on the same page as separate results, and the reader opened the ones that caught their eye and built a picture from the pieces.

That system was crude, but it gave everyone the same raw material. A bad story was one link among ten, and the reader could weigh it against the other nine. A good story worked the same way.

Reputation was a matter of position, meaning what ranked and in what order, and the whole industry that grew up around it, from SEO agencies to PR firms, worked on moving links up and down that list. If you wanted to know what the world saw when it searched your name, you searched your name, and the answer was the same for everyone.

2023: the reader stopped reading

Then ChatGPT arrived, and within a year the habit changed. Rather than typing a name and reading ten links, people started asking a question and reading one paragraph. The machine did the clicking and the skimming and the weighing, and it handed back a summary in a confident voice, with no links at all in the early versions.

For anyone whose name gets searched, this was the moment the rules changed, because a summary and a list are different things, and the difference matters more for reputation than for almost anything else.

A list shows the reader the raw material and lets them judge it. A summary makes the judgment for them.

When ten links included one hard story, the reader saw it as one of ten. When an AI folded the same story into a paragraph, the reader saw only the paragraph, so the story either became the main thing they learned about you or dropped out completely, and you had no way of knowing that.

2024 to 2025: the summary moved to the top of Google

Google watched this happen and responded by putting the summary on top of its own results. AI Overviews rolled out in 2024 and now sit above the links for more than one in five searches. In 2025 Google added AI Mode, a full conversational search, and according to figures compiled by Thrive, AI Mode passed one billion monthly users in 2026, with queries more than doubling every quarter since launch and only 6 to 8 percent of sessions ending in a click on any outside site.

The same source reports that people type questions into AI Mode that are three times longer than a normal search, and that they ask more and more follow-up questions in the same session. In plain terms, people no longer type a name and read a list. They ask the machine about you and keep asking. Then in January 2026 Apple placed Google's Gemini inside Siri, as SEO Sherpa records, which put an AI answer one voice command away on more than two billion devices.

So in the space of two years the list that people used to read became a paragraph that people now trust, and that paragraph reaches almost everyone. Whatever the machine says about you is now the first and often the only thing a buyer, an investor, a party official, or a journalist learns before they decide whether to pick up the phone.

2025 to 2026: the machine started getting it wrong in public

Here the stakes jumped again, because the paragraph is sometimes false, and the first court cases about false AI summaries are now working their way through the system.

Wolf River Electric, a solar installer in Minnesota, discovered in 2025 that Google's AI Overview was telling searchers the company had been sued by the state attorney general for misleading customers and hiding fees. The attorney general had sued four other solar companies, and Wolf River was never among them, but the summary said otherwise.

According to Government Technology, one customer read the summary and cancelled a 150,000 dollar contract, and the company is now suing Google in federal court for between 110 and 210 million dollars in damages.

In September 2026 a federal judge in Chicago let a second case go forward. Google's AI Overview had described an author and television producer as a convicted murderer serving life in prison for killing three women, when in fact he had a 1997 drug conviction that was later expunged after he worked as an FBI informant.

As Courthouse News reported, the court ruled that AI-generated statements can reasonably be read as statements of fact, and it allowed the defamation claim to proceed because the man had notified Google three times and the false summary kept appearing. Reputation X drew the practical lesson from the ruling, which is that the record of who was told what and when has become evidence.

These cases matter beyond the people in them. In 2015 a false claim about you needed a person to write it and a publication to run it, and both of them could be called, corrected, or sued. In 2026 a false claim can be assembled by a model from fragments of unrelated pages, shown to millions of people, and shown differently to each of them, and there is no editor at the other end of the phone.

What actually changed for reputation

Put the timeline together and five things have shifted, each one harder to manage than the last.

What we do about it

This is the job at Known, and AI search has made it both more urgent and more concrete. We start by asking the models the questions a buyer, an investor, or a party committee would ask, across ChatGPT, Gemini, Perplexity, and Google's AI Mode, and we keep asking on a schedule, because the answer moves.

When an answer is wrong we save it with the date and send a correction request to the platform, so that the mistake is on record and the model has a reason to update.

Then we fix the inputs. We have working relationships with editors at recognized business and political publications, and we run our own portfolio of media outlets, so we can place real coverage where the models look rather than waiting for a journalist to find you.

A profile that explains what a company does, an interview that puts a founder's thinking on the record, a piece of commentary that shows a politician understands the brief, each of these gives the model something accurate to build from, and over a few months the paragraph changes to match.

We do this for companies preparing a raise or a sale, for investors who need founders to take them seriously, for politicians facing selection or an election, and for public figures who have done the work but never had it written down where a machine could read it.

Where this goes next

The direction is set. Every year a larger share of first impressions will come from a paragraph a machine wrote, and the courts are only now starting to decide who answers for that paragraph when it is wrong.

Companies and public figures who wait for that to be settled will spend those years being described by whatever happens to be online about them, while the ones who treat the AI answer as a page of their own, something to be read, corrected, and fed with better material, will be described the way they intended.

If you would like to know what the models say about you right now, send us a name and we will run it through ChatGPT, Gemini, Perplexity, and Google's AI Mode the way a buyer, a board member or a vetting committee would, and within 24 hours you will have the full set of answers side by side, a note on what is wrong or missing, and a plain view of what it would take to change them.

Sources