I was on a call with a fintech CEO last month who'd just discovered his company's dirty laundry was being aired by an AI to every prospective client.
Not a journalist. Not a competitor's ad campaign. ChatGPT.
When potential investors typed his company name, the model summarized a lawsuit from 2019, a disgruntled Glassdoor review from 2021, and a forum thread speculating about insolvency. All three sources were outdated, two were legally resolved, but the AI presented them as "what you should know" before meeting the team.
His organic rankings were fine. His website was pristine. His PR team had buried the negative stories on page three of Google years ago. None of it mattered. The AI didn't care about page rank. It cared about what it could extract.
That's the new reputational battlefield. And traditional SEO is about as useful as a fire extinguisher in a flood.
Why Your Defenses Are Obsolete
The data is stark. About 30% of users now use both traditional search and generative AI when researching companies. Of those researching B2B services, roughly two-thirds start with Google, but nearly 30% go straight to ChatGPT, Gemini, or Perplexity.
Here's the critical difference: traditional search shows you a list of links and lets you decide. AI search decides for you. It synthesizes. It judges. And if negative information exists anywhere in its training data or live retrieval scope, it will fold that into the narrative—often without context, without recency, and without your ability to appeal.
Legacy SEO can't save you here. You can own the first page of Google and still get assassinated by an AI Overview that pulls from a toxic Reddit thread or a five-year-old lawsuit filing. You aren't fighting for position anymore. You're fighting for narrative control inside a black box.
What It Actually Costs
This isn't theoretical. A major fast-food chain took a 34.7 billion yen hit after a contamination issue spiraled into a reputational crisis. A cosmetics manufacturer lost over 5 billion yen from "white spot" damage that went viral. In the AI era, these numbers accelerate because the negative story doesn't just rank—it gets spoken by a trusted voice.
When an AI tells a user "this company has faced regulatory scrutiny," the user doesn't fact-check. They just don't buy. They don't sign. They don't show up to the meeting.
The Mercury Defense Framework
At Mercury, we approach this as Citation Engineering in reverse. Just as we engineer positive algorithmic authority for clients, we engineer defensive moats that make negative information structurally harder for AI models to cite, synthesize, and present as truth.
The framework has three stages. None of them involve hoping the problem goes away.
Stage One: Real-Time Monitoring (The Weekly Triad)
You cannot manage what you cannot see. And in the AI era, "seeing" means checking the machines that are doing the talking.
We run a weekly protocol across three tools: ChatGPT, Gemini, and Perplexity. We enter the company name, service names, and high-risk keyword combinations. We don't just look at the answer. We look at the sources the AI used to build the answer.
Perplexity is particularly valuable here because it shows its work—the actual URLs it pulled from. If a negative site is appearing as a source, we know exactly where the infection is coming from.
We score responses on a five-level positive scale. If a negative response appears, we trace it to the source within 24 hours. Monthly checks are too slow. By the time you notice, the AI has already told a hundred prospects.
Stage Two: Source Control (Building Your Algorithmic Immune System)
Once you know where the negative signal is coming from, you have two jobs: suppress the bad and fortify the good.
Entity Architecture & Structured Data
AI models trust what they can verify across multiple high-authority sources. So we build a fortress of primary information that the models must respect.
This starts with your own digital real estate. Every page on your site should have structured data markup—Schema.org implemented correctly: Organization type, Article type, FAQ type. This isn't a "nice to have." It's how you tell an AI, "This is who we are, this is what we do, and this data is authoritative."
We also ensure your official site contains the elements AI uses to judge credibility: comprehensive company profiles, management backgrounds, verified achievements, awards, IR information, and—crucially—fresh content. AI models weight recency heavily. A site that hasn't been updated in a year looks abandoned to an algorithm.
Third-Party Anchoring
AI doesn't trust your website alone. It wants consensus. So we strategically increase mentions across verified, high-trust external nodes: specialized media, academic citations, public institution databases, Wikipedia (where appropriate and accurate), and industry directories.
The goal is to make the AI's retrieval system hit so many positive, consistent, high-authority sources that the negative ones get statistically drowned out.
Owned Media & Primary Data
We publish proprietary research, case studies, and expert analysis that no competitor can replicate. AI models prioritize content with quantitative information—statistics, implementation results, survey data. By continuously publishing fresh, data-rich primary content, we occupy the information layer that the AI prefers to cite.
Stage Three: Reverse Citation Engineering (Surgical Removal)
Sometimes suppression isn't enough. The source needs to be cut off.
Reverse SEO
For negative content that ranks in traditional search, we deploy reverse SEO—not black-hat manipulation, but aggressive positive content deployment that pushes negative results off the first page. We aim to occupy 8 of the 10 first-page slots with positive or neutral information. The negative article doesn't need to disappear; it just needs to become invisible to anyone who doesn't scroll to page three.
Reverse AIO
This is the critical distinction. Reverse SEO targets Google rankings. Reverse AIO targets the AI's synthesis engine.
We identify the specific sources the AI is pulling from when it generates negative responses. Sometimes it's a defamatory article. Sometimes it's a forum thread. Sometimes it's a misinterpreted news piece. We then use legal deletion requests (where rights infringement is clear), content replacement, and source-priority degradation to make those sources less attractive to the model's retrieval system.
This is harder than reverse SEO because you're not fighting for position on a page. You're fighting to be excluded from a statistical model's reference set. But it's the only defense that actually works in the AI era.
Suggestion Pollution Control
AI models also reference search engine autocomplete suggestions. If typing your company name triggers "company name + scandal" or "company name + lawsuit," that signal feeds the AI's understanding of your brand.
We counter this by strategically increasing search volume for positive keyword combinations—"company name + awards," "company name + innovation," "company name + case study." Over time, this shifts the autocomplete landscape and the AI's associated semantic field.
The Five KPIs That Actually Matter
If you're running this internally or evaluating a vendor, track these five metrics monthly:
Table
KPI
How We Measure
Target
AI Positive Score
5-level evaluation across ChatGPT, Gemini, Perplexity
4.0+
Positive Keyword Search Volume
Google Search Console / trend data
150%+ YoY
Negative Suggestion Rate
Manual monthly check of autocomplete
Zero negative triggers
First-Page Positivity
Ratio of positive/neutral in top 10 Google results
80%+
AI Mention Velocity
Weekly count of positive AI citations
Month-over-month growth
If your vendor is only tracking traditional rankings, they're fighting the last war.
The Organizational Reality
This isn't a one-time fix. It's a permanent operational function.
Internally, it requires coordination between PR (media relations), Marketing (content and data), and Legal (deletion requests and compliance). Weekly standups. Shared dashboards. Rapid decision-making.
If you lack internal bandwidth, this is where external specialization matters. But be careful who you hire. The market is full of agencies selling "AI reputation management" that is just old SEO with a new label.
At Mercury, we don't do traditional SEO. We do Citation Engineering—both offensive (building your algorithmic authority) and defensive (protecting it from contamination). We run the weekly triad checks. We build the entity architecture. We execute the reverse citation protocols. And we measure everything against the five KPIs above.
The Cost of Inaction vs. Action
Reverse SEO and AIO measures range widely depending on severity— from basic suggestion optimization to comprehensive legal-backed content suppression. But the cost of doing nothing is a slow bleed of trust that doesn't show up in your traffic reports until the pipeline has already dried up.
If negative information is being synthesized by AI models that your prospects consult before every major decision, you are losing deals you never knew you were in the running for.
The First Step
If you haven't checked what the AI says about you lately, do it now. Open ChatGPT, Gemini, and Perplexity. Type your company name. Read the response. Check the sources.
If you don't like what you see, the time to start building your defense was yesterday. The second-best time is today.
— James, Mercury Technology SolutionsLearn more at www.mtsoln.comHong Kong, May 2026
