This article was AI-generated as part of an experimental historical-content project. The date reflects the period being analyzed rather than the date the article was originally written.
Most coverage of Google’s September Bard update led with Extensions, which let Bard pull information from Gmail, Docs, Drive, Maps, YouTube, and Google Flights and hotels. The more revealing change was a small “G” icon.
On September 19, Google said Bard’s “Google it” button would now help double-check its own answers. Click it, and Bard reads its response back and evaluates whether there is content across the web to substantiate it. Where a statement can be evaluated, the phrase is highlighted, and you can click through to supporting or contradicting information found by Search.
Google presented this as a convenience. I read it as a product admission: a fluent answer is not yet a trusted answer, and the trust still has to be borrowed from the open web.
Reuters reported the same day that Bard drew 183 million visits in August, about 13% of ChatGPT’s traffic, according to Similarweb. Google is the one chasing, and the feature it chose to emphasize was verification. That hints at where it thinks its advantage sits: it has the index to check against.
A double-check tests agreement, not truth
This is the part communications teams should sit with. The “Google it” check does not establish whether a statement is true. It establishes whether the web agrees with it.
Usually that is a sensible proxy. For questions about a specific company or executive, it can fail in two ways. If the web is thin about you, a statement may simply not be evaluable, and the user is left with an unverified sentence and no flag. If the web is wrong about you (an outdated title, an old dispute repeated by aggregators, a figure from an article that was later corrected), the check may appear to substantiate the error, because the error is what the indexed sources say.
Put simply, the fact-check inherits your search results. Whatever has accumulated about you online becomes the reference text the chatbot grades itself against.
Reputation teams have long dealt with stale information that keeps ranking because it was written once and copied many times. What changes is its role. A stale article on page two used to be a stale article on page two. Inside a verification loop, it can become the evidence that makes a wrong answer look checked. Google’s generative search experiment added publish dates to supporting links in August for similar reasons: provenance is starting to carry real weight.
What to do with this
None of this calls for “optimizing for Bard.” It is basic hygiene with a second job.
The core facts about a company (leadership, headquarters, what it actually does, key dates) should be easy to find, consistent across your own pages, and reflected in credible third-party sources. Inconsistency is exactly what makes verification ambiguous.
Corrections also need to travel. If a widely cited article contains an error, a note on your own site helps less than getting the original source, or a more authoritative one, to state the correct fact.
Bard is still labeled an experiment and trails ChatGPT on usage, but the habit of asking a chatbot about a company is forming either way.
Google built this button because it knows generated answers need a path back to evidence. That instinct is right. It also means the evidence about your company, current or stale, is now doing double duty.