The next frontier of online reputation risk

As generative AI turns into extra embedded in search and content material experiences, it’s additionally rising as a brand new supply of misinformation and reputational hurt. 

False or deceptive statements generated by AI chatbots are already prompting authorized disputes – and elevating contemporary questions on legal responsibility, accuracy, and on-line status administration.

When AI turns into the supply of defamation

It’s unsurprising that AI has grow to be a brand new supply of defamation and on-line status injury. 

As an SEO and status knowledgeable witness, I’ve already been approached by litigants concerned in circumstances the place AI techniques produced libelous statements.

That is uncharted territory – and whereas options are rising, a lot of it stays new floor.

Actual-world examples of AI-generated defamation

One consumer contacted me after Meta’s Llama AI generated false, deceptive, and defamatory statements a few distinguished particular person. 

Early analysis confirmed that the particular person had been concerned in – and prevailed in – earlier defamation lawsuits, which had been reported by information shops. 

Some detractors had additionally criticized the person on-line, and discussions on Reddit included inaccurate and inflammatory language. 

But when the AI was requested in regards to the particular person or their status, it repeated these vanquished claims, added new warnings, and projected assertions of fraud and untrustworthiness.

In one other case, a consumer focused by defamatory weblog posts discovered that just about any immediate about them in ChatGPT surfaced the identical false claims. 

The important thing concern: even when the court docket orders the unique posts eliminated, how lengthy will these defamatory statements persist in AI responses?

Google Traits exhibits that there was a big spike in searches associated to defamation communicated by way of AI chatbots and AI-related on-line status administration:

Fabricated tales and real-world hurt

In different circumstances revealed by lawsuit filings, generative AI has apparently fabricated completely false and damaging content material about folks out of skinny air. 

In 2023, Jonathan Turley, the Shapiro Professor of Public Curiosity Regulation at George Washington College, was falsely reported to have been accused of sexual harassment – a declare that was by no means made, on a visit that by no means occurred, whereas he was at a school the place he by no means taught. 

ChatGPT cited a Washington Submit article that was by no means written as its supply.

In September, former FBI operative James Keene filed a lawsuit towards Google after its AI falsely claimed he was serving a life sentence for a number of convictions and described him because the assassin of three ladies. 

The swimsuit additionally alleges that these false statements have been doubtlessly seen by tens of thousands and thousands of searchers.

Generative AI can fabricate tales about folks – that’s the “generative” a part of “generative AI.” 

After receiving a immediate, an AI chatbot analyzes the enter and produces a response primarily based on patterns realized from massive volumes of textual content. 

So it’s no shock that AI solutions have at instances included false and defamatory content material about people.

Enhancements and remaining challenges

Over the previous two years, AI chatbots have proven enchancment in dealing with biographical details about people.

Probably the most distinguished chatbot corporations appear to have centered on refining their techniques to higher handle queries involving folks and correct names.

Consequently, the technology of false info – or hallucinations – about people appears to have declined considerably.

AI chat suppliers have additionally begun incorporating extra disclaimer language into responses about folks’s biographical particulars and reputations.

These typically embody statements noting:

• Restricted info.
• Uncertainty about an individual’s identification.
• The dearth of impartial verification.

It’s unclear how a lot such disclaimers really shield towards false or damaging assertions, however they’re no less than preferable to offering no warning in any respect.

In a single occasion, a consumer who was allegedly defamed by Meta’s AI had their counsel contact the corporate immediately.

Meta reportedly moved rapidly to handle the difficulty – and will have even apologized, which is sort of exceptional in issues of company civil legal responsibility.

At this stage, the best reputational dangers from AI are much less about outright fabrications.

The extra urgent threats come from AI techniques:

• Misconstruing supply materials to attract inaccurate conclusions.
• Repeating others’ defamatory claims.
• Exaggerating and distorting true details in deceptive methods.

Authorized legal responsibility and Part 230

As a result of the regulation round AI-generated libel remains to be quickly growing, there may be little authorized precedent defining how liable corporations is perhaps for defamatory statements produced by their AI chatbots.

Some argue that Part 230 of the Communications Decency Act may defend AI corporations from such legal responsibility.

The reasoning is that if on-line platforms are largely immune from defamation claims for third-party content material they host, then AI techniques needs to be equally protected since their outputs are derived from third-party sources.

Nevertheless, derived is much from quoted or reproduced – it implies a significant diploma of originality.

If legislators already believed AI output was protected beneath Part 230, they doubtless wouldn’t have proposed a 10-year moratorium on imposing state or native restrictions on synthetic intelligence fashions, techniques, and decision-making processes.

That moratorium was initially included in President Trump’s funds reconciliation invoice, H.R.1 – nicknamed the “One Massive Lovely Invoice Act” – however was in the end dropped when the regulation was signed on July 4, 2025.

AI’s rising position in status administration

The rising prominence of AI-generated solutions – similar to Google’s AI Overviews – is making details about folks’s backgrounds and reputations each extra seen and extra influential. 

As these techniques grow to be more and more correct and reliable, it’s not a stretch to say that the general public will likely be extra inclined to consider what AI says about somebody – even when that info is fake, deceptive, or defamatory.

AI can also be taking part in a bigger position in background checks. 

For instance, Checkr has developed a customized AI that searches for and surfaces doubtlessly adverse or defamatory details about people – findings that might restrict an individual’s employment alternatives with corporations utilizing the service. 

Whereas main AI suppliers similar to Google, OpenAI, Microsoft, and Meta have applied guardrails to cut back the unfold of defamation, providers like Checkr are much less more likely to embody caveats or disclaimers. 

Any defamatory content material generated by such techniques might due to this fact go unnoticed by these it impacts.

At current, AI is more than likely to supply defamatory statements when the net already accommodates defamatory pages or paperwork. 

Eradicating these supply supplies normally corrects or eliminates the false info from AI outputs. 

However as AI techniques more and more “keep in mind” prior responses – or cache info to save lots of on processing – eradicating the unique sources might not be sufficient to erase defamatory or misguided claims from AI-generated solutions.

What could be finished about AI defamation?

One key option to handle defamation showing in AI platforms is to ask them on to right or take away false and damaging statements about you. 

As famous above, some platforms – similar to Meta – have already taken motion to take away content material that appeared libelous. 

(Sarcastically, it might now be simpler to get Meta to delete dangerous materials from its Llama AI than from Fb.)

These corporations could also be extra responsive if the request comes from an lawyer, although additionally they seem keen to behave on reviews submitted by people.

Right here’s how you can contact every main AI supplier to request the removing of defamatory content material:

Meta Llama

Use the Llama Developer Feedback Form or e mail [email protected] to report or request removing of false or defamatory content material.

ChatGPT

In ChatGPT, you may report problematic content material immediately throughout the chat interface. 

On desktop, click on the three dots within the upper-right nook and choose Report from the dropdown menu. 

On cellular or different gadgets, the choice might seem beneath a unique menu.

AI Overviews and Gemini

There are two methods to report content material to Google. 

You possibly can report content for legal reasons. (Click on See extra choices to pick out Gemini, or throughout the Gemini desktop interface, use the three dots beneath a response.)

Nevertheless, Google sometimes gained’t take away content material via this route except you could have a court docket order, because it can not decide whether or not materials is defamatory.

Alternatively, you may ship suggestions immediately. 

For AI Overviews, click on the three dots on the suitable aspect of the consequence and select Suggestions. 

From Gemini, click on the thumbs-down icon and full the suggestions kind. 

Whereas this method might take time, Google has beforehand diminished visibility of dangerous or deceptive info via delicate suppression – much like its method with Autocomplete. 

When submitting suggestions, clarify that:

• You aren’t a public determine.
• The AI Overview unfairly highlights adverse materials.
• You’ll admire Google limiting its show even when the supply pages stay on-line.

Bing AI Overview and Microsoft Copilot

As with Google, you may both ship suggestions or report a priority. 

In Bing search outcomes, click on the thumbs-down icon beneath an AI Overview to start the suggestions course of. 

Within the Copilot chatbot interface, click on the thumbs-down icon beneath the AI-generated response.

When submitting suggestions, describe clearly – and politely – how the content material about you is inaccurate or dangerous.

For authorized removing requests, use Microsoft’s Report a Concern kind. 

Nevertheless, this route is unlikely to succeed with out a court docket order declaring the content material unlawful or defamatory.

Perplexity

To request the removing of details about your self from Perplexity AI, e mail [email protected] with the related particulars.

Grok AI

You possibly can report a problem inside Grok by clicking the three dots beneath a response. Authorized points may also be reported via xAI. 

In line with xAI’s privateness coverage:

• “Please notice that we can not assure the factual accuracy of Output from our fashions. If Output accommodates factually inaccurate private info referring to you, you may submit a correction request and we’ll make cheap efforts to right this info – however as a result of technical complexity of our fashions, it might not be possible for us to take action.”

To submit a correction request, go to https://xai-privacy.relyance.ai/.

Extra approaches to addressing status injury in AI

If contacting AI suppliers doesn’t totally resolve the difficulty, there are different steps you may take to restrict or counteract the unfold of false or damaging info.

Take away adverse content material from originating sources

Exterior of the reducing cases of defamatory or damaging statements produced by AI hallucinations, most dangerous content material is gathered or summarized from present on-line sources. 

Work to take away or modify these sources to make it much less doubtless that AIs will floor them in responses. 

Persuasion is step one, the place potential. For instance:

• Add a press release to a information article acknowledging factual errors.
• Notice {that a} court docket has dominated the content material false or defamatory.

These can set off AI guardrails that stop the fabric from being repeated. 

Disclaimers or retractions can also cease AI techniques from reproducing adverse info.

Overwhelm AI with optimistic and impartial info

Proof means that AIs are influenced by the amount of constant info accessible. 

Publishing sufficient correct, optimistic, or impartial materials about an individual can shift what an AI considers dependable. 

If most sources mirror the identical biographical particulars, AI fashions might favor these over remoted adverse claims. 

Nevertheless, the brand new content material should seem on respected websites which are equal to or superior in authority to the place the adverse materials was revealed – a problem when the dangerous content material originates from main information shops, authorities web sites, or different credible domains.

Displace the adverse info within the search engine outcomes

Main AI chatbots source some of their information from search engines. 

Primarily based on my testing, the complexity of the question determines what number of outcomes an AI might reference, starting from the primary 10 listings to a number of dozen or extra. 

The implication is obvious: if you happen to can push adverse outcomes additional down in search rankings – past the place the AI sometimes appears to be like – these gadgets are much less more likely to seem in AI-generated responses.

This can be a traditional on-line status administration methodology: using commonplace website positioning methods and a community of on-line property to displace adverse content material in search outcomes. 

Nevertheless, AI has added a brand new layer of issue. 

ORM professionals now want to find out how far again every AI mannequin scans outcomes to reply questions on an individual or matter. 

Solely then can they understand how far the damaging outcomes should be pushed to “clear up” AI responses.

Prior to now, pushing adverse content material off the primary one or two pages of search outcomes offered about 99% reduction from its affect. 

As we speak, that’s typically not sufficient. 

AI techniques might pull from a lot deeper within the search index – that means ORM specialists should suppress dangerous content material throughout a wider vary of pages and associated queries. 

As a result of AI can conduct a number of, semantically associated searches when forming solutions, it’s important to check numerous key phrase combos and clear adverse gadgets throughout all related SERPs.

Obfuscate by launching personas that share the identical title

Utilizing personas that “coincidentally” share the identical title as somebody experiencing status issues has lengthy been an occasional, last-resort technique. 

It’s most related for people who’re uncomfortable creating extra on-line media about themselves – even when doing so may assist counteract unfair, deceptive, or defamatory content material. 

Sarcastically, that reluctance typically contributes to the issue: a weak on-line presence makes it simpler for somebody’s status to be broken.

When a reputation is shared by a number of people, AI chatbots seem to tread extra fastidiously, typically avoiding particular statements after they can’t decide who the data refers to. 

This tendency could be leveraged. 

By creating a number of well-developed on-line personas with the identical title – full with legitimate-seeming digital footprints – it’s potential to make AIs much less sure about which particular person is being referenced. 

That uncertainty can stop them from surfacing or repeating defamatory materials.

This methodology shouldn’t be with out issues. 

Individuals more and more use each AI and conventional search instruments to seek out private info, so including new identities dangers confusion or unintended publicity. 

Nonetheless, in sure circumstances, “clouding the waters” with credible alternate personas could be a sensible option to scale back or dilute defamatory associations in AI-generated responses.

Previous legal guidelines, new dangers

A hybrid method combining the strategies described above could also be essential to mitigate the hurt skilled by victims of AI-related defamation.

Some types of defamation have at all times been troublesome – and generally unimaginable – to handle via lawsuits. 

Litigation is pricey and may take months or years to yield reduction. 

In some circumstances, pursuing a lawsuit is additional sophisticated by skilled or authorized constraints. 

For instance, a physician in search of to sue a affected person over defamatory statements may violate HIPAA by disclosing figuring out info, and attorneys might face comparable challenges beneath their respective bar affiliation ethics guidelines.

There’s additionally the chance that defamation lengthy buried in search outcomes – or barred from litigation by statutes of limitation – may abruptly resurface via AI chatbot responses. 

It could finally result in attention-grabbing case regulation, arguing that an AI-generated response constitutes a “new publication” of defamatory content material, doubtlessly resetting the restrictions on these claims.

One other potential resolution, albeit a distant one, can be to advocate for brand new laws that protects people from adverse or false info disseminated via AI techniques. 

Different areas, similar to Europe, have established privateness legal guidelines, together with the “Proper to be Forgotten,” that give people extra management over their private info. 

Comparable protections can be priceless in the USA, however they continue to be unlikely given the enduring drive of Part 230, which continues to defend massive tech corporations from legal responsibility for on-line content material.

AI-driven reputational hurt stays a quickly evolving discipline – legally, technologically, and strategically. 

Anticipate additional developments forward as courts, lawmakers, and technologists proceed to grapple with this rising frontier.