AI Search Crisis Management: Protecting Brand Reputation

It is 6:47 AM on a Tuesday. Your phone is vibrating non-stop. Your VP of Sales has forwarded a screenshot from a prospect: ChatGPT just told them your platform has a critical security vulnerability that exposes customer payment data. The prospect wants out of the deal. The problem? There is no vulnerability. There never was. An outdated forum post from three years ago discussed a hypothetical risk that was patched before it ever reached production, and now an AI model has woven it into a confident, authoritative-sounding answer that is being served to thousands of users every day.

Your stomach drops. Your brand is being damaged at scale, and there is no “report this result” button, no webmaster tools appeal, no ad to run above it. Welcome to the new frontier of AI reputation management, where a single hallucinated citation can unravel years of trust-building in hours.

This guide is your crisis playbook. Whether the threat is real or fabricated, whether the damage is contained or spiraling, you will learn exactly how to monitor, detect, respond, recover, and prevent reputation crises in AI search. Step by step. Hour by hour. With frameworks you can deploy before your morning coffee gets cold.

Why AI Reputation Crises Are Different From Anything You Have Faced Before

Traditional reputation crises follow patterns that PR professionals have managed for decades. A bad review appears on G2. A disgruntled employee posts on Glassdoor. A journalist writes a critical article. In each case, the damage is visible, traceable, and confined to a specific platform with its own dispute mechanisms.

AI reputation management requires an entirely different mindset because the mechanics of damage are fundamentally different.

The Amplification Problem

When ChatGPT, Perplexity, Gemini, or Claude presents negative information about your brand, it does not present it as one opinion among many. It presents it as synthesized fact. There is no comment section where your team can post a correction. There is no star rating that averages out over time. The AI delivers a single, confident narrative, and the user has no reason to question it.

Worse, AI models pull from multiple sources and blend them together. A mildly negative product review, a competitor’s comparison page that positions you unfavorably, and an old support thread about a bug that was fixed two years ago can combine into a response that makes your product sound fundamentally broken.

The Invisibility Problem

In traditional reputation management, you can Google yourself and see the problem. You can set up alerts. You can monitor review sites. With AI search, the damaging narrative might only surface in response to specific prompts that you would never think to search for. A prospect asking “Is [Your Product] secure enough for healthcare data?” might trigger a completely different response than “Best healthcare SaaS platforms,” and you would never know unless someone told you.

This is what makes a brand crisis AI scenario so dangerous. The crisis can be happening at scale, silently, without any of the traditional warning signs appearing in your monitoring dashboards.

The Persistence Problem

When a negative article ranks on Google, you can work to outrank it. When bad information gets embedded in an AI model’s training data or retrieval pipeline, the correction process is far less direct. AI models update on their own schedules. Retrieval-augmented generation systems pull from indexes that refresh at varying intervals. A correction you publish today might not influence AI responses for weeks or months, and during that entire period, every user asking the relevant question gets the wrong answer.

Related: Why Your SaaS Isn’t Showing Up in AI Search Results

The Five-Phase Crisis Framework: Monitor, Detect, Respond, Recover, Prevent

Effective AI reputation management follows a continuous cycle. Each phase feeds into the next, and skipping any one of them leaves you exposed.

Phase	Objective	Timeframe	Key Activities
Monitor	Establish visibility into AI outputs	Ongoing	Automated querying, sentiment tracking, citation auditing
Detect	Identify threats before they escalate	Real-time to 24 hours	Alert systems, threshold triggers, pattern recognition
Respond	Contain damage and correct the narrative	First 2-48 hours	Stakeholder comms, content publishing, platform outreach
Recover	Rebuild positive positioning	1-12 weeks	Content amplification, authority building, citation recapture
Prevent	Harden against future crises	Ongoing	Content moats, entity consistency, proactive positioning

This framework borrows from traditional crisis communication theory, particularly the Situational Crisis Communication Theory (SCCT) developed by W. Timothy Coombs, but adapts it specifically for the unique dynamics of AI-generated content. The principles are the same: take responsibility where appropriate, communicate transparently, and rebuild trust through consistent action. The tactics, however, are entirely new.

Phase 1: Building Your AI Monitoring System

You cannot manage what you cannot see. Before any crisis hits, you need a monitoring system that gives you visibility into what AI models are saying about your brand.

Automated Query Monitoring

Set up a systematic process for querying major AI platforms about your brand on a regular cadence. This is not something you do once. It is something that runs daily or, at minimum, weekly.

Core query categories to monitor:

• Direct brand queries: “What is [Your Brand]?” / “Tell me about [Your Brand]”
• Comparison queries: “[Your Brand] vs [Competitor]” / “Best alternatives to [Your Brand]”
• Reputation queries: “Is [Your Brand] reliable?” / “[Your Brand] reviews” / “[Your Brand] problems”
• Category queries: “Best [your category] software for [use case]”
• Negative signal queries: “[Your Brand] security issues” / “[Your Brand] complaints” / “[Your Brand] outage”

For each query, log the full response, the date, the platform, and any sources cited. Over time, this creates a dataset that lets you spot trends before they become crises.

Monitoring Tools and Approaches

Currently, no single tool provides comprehensive reputation monitoring ChatGPT and other AI platforms. You will likely need a combination:

• Custom scripts that query AI APIs (OpenAI, Anthropic, Perplexity) on a schedule and store responses for comparison
• Brand monitoring platforms like Mention, Brandwatch, or Brand24 that track traditional web mentions (which feed into AI training data)
• Citation tracking tools that monitor when and how AI models reference your brand
• Source monitoring focused on the specific pages and domains that AI models frequently cite in your industry

The goal is to build a baseline understanding of what AI models say about you when nothing is wrong, so that you can immediately recognize when something changes.

Related: AI Search Analytics: Tracking ChatGPT and Perplexity Traffic in GA4

Sentiment Scoring

Not every mention is a crisis. You need a way to categorize and score the sentiment of AI-generated responses about your brand. A simple three-tier system works well for most organizations:

• Green (Positive/Neutral): Brand mentioned favorably or factually, no corrective action needed
• Yellow (Concerning): Brand mentioned with minor inaccuracies, outdated info, or lukewarm positioning
• Red (Critical): Brand associated with security failures, data breaches, ethical violations, or significantly false claims

Any response that hits “Red” triggers your crisis response protocol immediately. “Yellow” responses get queued for corrective content within the week.

Phase 2: Detection Protocols That Catch Problems Early

Monitoring gives you data. Detection gives you speed. The difference between a manageable reputation issue and a full-blown brand crisis AI disaster often comes down to how quickly you notice the problem.

Early Warning Signals

Watch for these indicators that a crisis may be forming:

• Sudden changes in AI response sentiment for queries that were previously stable
• New negative sources appearing in AI citations that were not there before
• Customer support tickets or sales objections referencing information that sounds like it came from an AI
• Social media posts where users share AI-generated responses about your brand
• Competitor content being published that targets your known vulnerabilities

Detection Automation

Build alert triggers that notify your team immediately when thresholds are crossed:

• Sentiment shift alerts: If a monitored query drops from Green to Yellow or Red
• New source alerts: If AI responses begin citing a source that was not previously in the citation set
• Volume alerts: If the number of customer inquiries referencing AI-sourced information spikes
• Competitor alerts: If a competitor publishes content specifically designed to influence AI responses about your brand

The faster you detect, the faster you respond. And in negative citation management, speed is everything.

Phase 3: The Crisis Response Playbook (Hour by Hour)

When a crisis is detected, your response must be immediate, structured, and disciplined. Panic-driven reactions almost always make things worse. Here is the hour-by-hour protocol.

The First 2 Hours: Assess and Contain

This is the most critical window. What you do in the first 120 minutes determines whether the crisis escalates or starts to stabilize.

Minute 0-30: Assessment

• Confirm the negative information exists across multiple AI platforms, not just one
• Determine whether the information is factually false, misleading, outdated, or accurate but damaging
• Identify the likely source(s) feeding the negative narrative to AI models
• Assess the reach: How many users are likely encountering this information?

Minute 30-60: Internal Mobilization

• Brief your crisis team: PR lead, legal counsel, product/engineering lead, CEO or founder
• Establish a single source of truth document that all external communications will reference
• Assign clear roles: Who owns the narrative? Who handles customer inquiries? Who manages content publishing?
• Do not publish anything externally yet. Premature public statements often contain errors that create secondary crises

Minute 60-120: Immediate Content Response

• Draft a factual correction statement optimized for AI retrieval
• Publish an authoritative page on your own domain addressing the specific claim (blog post, knowledge base article, or official statement)
• Update any existing pages that AI models are citing to ensure they contain current, accurate information
• If the negative information stems from a specific third-party source, initiate outreach to request correction

The First 24 Hours: Correct and Communicate

With the immediate assessment complete, the next 22 hours focus on correction and controlled communication.

Content publishing priorities (hours 2-12):

1. Official response page on your domain, written for both humans and AI retrieval. Use clear, factual language. Include schema markup to help AI models identify this as an authoritative correction
2. Updated FAQ entries that directly address the claims being made
3. Technical documentation updates if the crisis involves product claims (security, performance, reliability)
4. Third-party outreach to authoritative sites in your space, requesting they publish or update content reflecting the accurate information

Stakeholder communication priorities (hours 2-24):

1. Existing customers: Direct email if the crisis is significant enough that they might encounter the AI-generated claims. Be transparent, be factual, and provide clear evidence
2. Sales team: Equip them with talk tracks and supporting documentation to handle prospect objections
3. Partners and investors: Brief them before they hear about it elsewhere
4. Media contacts: Only if the crisis has reached traditional media. Otherwise, proactive media engagement can amplify a problem that users might not have noticed

Related: Content Optimization for LLMs: Writing for AI and Humans

The First Week: Amplify and Monitor

The initial fire is contained. Now you need to ensure the correction takes hold across AI platforms.

Days 2-7 action plan:

• Publish supporting content on third-party platforms (industry publications, partner blogs, review responses) that reinforces the accurate narrative
• Engage authoritative voices in your industry to comment on or share factual information about the situation
• Monitor AI responses daily for the specific queries that triggered the crisis. Track whether corrections are being picked up
• Document everything. Every response you publish, every outreach email you send, every change in AI output. This becomes your playbook for the next crisis and your evidence trail if legal action becomes necessary
• Conduct a source audit. Identify every web page that contributed to the negative AI narrative and create a plan to address each one

Phase 4: Recovery and Positive Content Amplification

Correcting false information is necessary but insufficient. After a brand crisis AI event, you need to actively rebuild the positive narrative that AI models associate with your brand.

The Content Saturation Strategy

The principle is straightforward: overwhelm the negative signal with positive signal. AI models weigh recency, authority, and consistency. If the corrective and positive content you publish after a crisis is more recent, more authoritative, and more consistent than the negative content, AI models will gradually shift their responses.

Positive content amplification tactics:

• Publish original research that positions your brand as a thought leader. Data-driven content is heavily weighted by AI models because it provides unique, citable information
• Secure third-party endorsements from recognized authorities in your space. Guest posts, podcast appearances, analyst reports, and industry awards all create positive signals
• Update and expand your knowledge base with comprehensive, well-structured content that directly addresses the topics where your brand was misrepresented
• Build case studies and customer success stories that provide concrete evidence contradicting the negative narrative
• Strengthen your entity presence by ensuring your brand information is consistent across Wikipedia, Crunchbase, LinkedIn, G2, and other authoritative reference sources

Citation Recapture

After a crisis, audit which queries now produce unfavorable results and create a targeted content plan for each one. This is negative citation management in its most proactive form.

For each problematic query:

1. Identify what the AI model is currently citing as its source
2. Create content that is more authoritative, more comprehensive, and more recent than that source
3. Ensure your content is structured for AI retrieval (clear headings, factual statements, schema markup)
4. Build backlinks and social signals to that content to establish authority
5. Monitor whether the AI response shifts over time

Related: The AI Citation Pyramid: Building Authority AI Agents Trust

Phase 5: Prevention Strategies That Make You Crisis-Resistant

The best crisis is the one that never happens. Prevention in AI reputation management means building a content and authority foundation so strong that negative signals struggle to break through.

Build a Content Moat

A content moat is a body of authoritative, comprehensive, and frequently updated content that dominates the information landscape for queries related to your brand and category.

Content moat components:

• Definitive guides for every major topic in your space, positioned on your domain
• Regularly updated comparison pages that fairly evaluate your product alongside competitors (AI models reward balanced, factual comparisons)
• A robust FAQ and knowledge base that preemptively answers every question a user or AI model might ask about your product
• Original data and research that cannot be found anywhere else, making your site an essential source for AI models
• Technical documentation that is clear, comprehensive, and optimized for AI crawlers

Entity Consistency

AI models build entity profiles for brands, assembling information from dozens of sources into a coherent understanding. If your brand information is inconsistent across platforms, AI models are more likely to surface inaccurate or outdated information.

Ensure consistency across:

• Your website’s about page, team page, and product descriptions
• LinkedIn company profile and employee profiles
• Crunchbase, G2, Capterra, and other SaaS directories
• Wikipedia (if applicable) and Wikidata
• Social media bios and descriptions
• Press releases and media kit materials
• Your llms.txt file and structured data markup

Every inconsistency is a crack where misinformation can seep in. Seal them all.

Related: llms.txt Implementation: Complete Guide for SaaS Companies

Proactive Reputation Positioning

Do not wait for a crisis to think about how AI models perceive your brand. Proactively shape that perception:

• Regularly query AI models about your brand and competitors to stay informed
• Publish thought leadership content that establishes your team as experts
• Maintain active relationships with industry analysts and journalists whose content carries authority with AI models
• Respond to every customer review on public platforms, positive and negative, with professionalism and substance
• Contribute to open-source projects, industry standards, and professional communities that build your brand’s authority profile

Content Suppression Tactics for Negative AI Citations

Sometimes the best defense is reducing the visibility of content that is feeding negative AI responses. This is not about censorship. It is about ensuring that outdated, inaccurate, or misleading content does not continue to distort AI-generated narratives about your brand.

Direct Source Remediation

• Contact website owners hosting inaccurate content and request corrections or removal. Many publishers will update factual errors when presented with evidence
• Submit correction requests to review platforms where outdated negative reviews are influencing AI citations
• Use legal channels (sparingly) for content that is genuinely defamatory. Cease-and-desist letters should be a last resort, as they can backfire publicly, but they are appropriate when content is provably false and damaging

Indirect Suppression Through Volume

You cannot remove content from the internet, but you can reduce its relative prominence in the information ecosystem:

• Publish higher-authority content on the same topics, optimized for AI retrieval
• Build backlinks to your corrective and positive content to increase its domain authority relative to the negative source
• Encourage positive reviews and testimonials that create a more balanced overall sentiment profile
• Engage in strategic guest posting on authoritative sites to create multiple positive touchpoints

AI Platform-Specific Actions

Each major AI platform has emerging mechanisms for addressing inaccuracies, though these are still developing:

• OpenAI: Feedback mechanisms within ChatGPT allow users to flag inaccurate responses. While individual flags may have limited impact, documented patterns of inaccuracy can influence model updates
• Perplexity: Because Perplexity uses real-time search, updating source content often reflects in responses more quickly than with other platforms
• Google Gemini: Connected to Google’s broader search ecosystem, meaning traditional SEO improvements can influence Gemini’s responses
• Anthropic Claude: Focuses heavily on cited sources, so improving the quality and accuracy of content Claude retrieves can shift responses

Related: Perplexity AI Optimization: Dominate Citations

Stakeholder Communication During an AI Reputation Crisis

How you communicate during a crisis matters as much as what you do to fix the underlying problem. Different stakeholders need different messages at different times.

Internal Communications

Your team hears about the crisis first. Do not let employees learn about a reputation issue from customers or social media. Internal communication should be:

• Immediate: Brief your leadership team within the first hour
• Factual: Share exactly what is happening without speculation or blame
• Directive: Give every team member clear guidance on what to say (and what not to say) if asked
• Updated regularly: As the situation evolves, keep your team informed so they are never caught off guard

Customer Communications

Not every AI reputation issue warrants customer communication. Use this threshold: If a customer is likely to encounter the negative AI-generated information in the normal course of evaluating or using your product, communicate proactively.

When you do communicate:

• Lead with facts, not apologies (especially if the information is false)
• Provide specific evidence that contradicts the inaccurate claims
• Give customers a clear point of contact for questions
• Follow up once the situation is resolved with a summary of actions taken

Investor and Board Communications

Investors and board members need to understand three things:

1. What happened: The specific nature of the AI-generated reputation issue
2. What the business impact is: Pipeline affected, deals at risk, customer churn potential
3. What you are doing about it: Your response plan, timeline, and expected outcomes

Be candid about the limitations of negative citation management in AI systems. Sophisticated investors understand that this is an emerging challenge without perfect solutions, and they will respect transparency more than false assurances.

Media Communications

For most AI reputation crises, do not proactively engage the media. The majority of AI-generated reputation issues are invisible to journalists, and drawing media attention to them can amplify the problem significantly.

The exception: If the crisis has already been picked up by media outlets or is trending on social media, prepare a brief, factual media statement. Keep it focused on what is true, what steps you are taking, and where people can find accurate information.

Recovery Timeline: What Realistic Progress Looks Like

Setting realistic expectations is critical. AI reputation management recovery does not happen overnight, and promising stakeholders a quick fix will damage your credibility when the timeline extends.

Typical Recovery Milestones

Timeframe	What to Expect
Week 1	Crisis contained. Corrective content published. Internal and customer communications complete. AI responses may not yet reflect corrections.
Weeks 2-4	Corrective content begins appearing in AI retrieval results for real-time platforms (Perplexity). Slower platforms may still show old information. Positive content amplification underway.
Weeks 4-8	Measurable shift in AI response sentiment for monitored queries. Some platforms fully updated, others lagging. Customer confidence stabilizing.
Weeks 8-16	Majority of AI platforms reflecting corrected information. Positive content amplification showing results. Brand sentiment metrics approaching or exceeding pre-crisis baseline.
Months 4-6	Full recovery for most crises. Residual negative mentions may persist in edge-case queries but no longer dominate primary brand queries. Prevention systems strengthened based on crisis learnings.

Factors That Accelerate Recovery

• Strong pre-crisis authority: Brands with established content moats and high domain authority recover faster because AI models already trust their content
• Speed of initial response: The faster you publish corrective content, the sooner it enters the AI retrieval pipeline
• Quality of corrective content: Well-structured, authoritative, evidence-based content is picked up faster than thin reactive posts
• Third-party validation: Corrections corroborated by independent authoritative sources carry more weight than self-published corrections alone

Factors That Slow Recovery

• Training data embedding: If negative information has been baked into a model’s training data (not just its retrieval pipeline), correction requires a model update cycle, which you cannot control
• Multiple negative sources: A single inaccurate source is easier to address than a pattern of negative content across many domains
• Ongoing negative activity: If a competitor or bad actor is continuously publishing negative content, recovery becomes an ongoing battle rather than a one-time correction
• Low pre-crisis authority: Brands with thin content profiles have less positive signal to counteract negative information

Related: How We Increased AI Citations by 600% in 90 Days

Putting It All Together: Your AI Crisis Management Checklist

Before a crisis hits, make sure you have these systems and resources in place:

Monitoring Infrastructure:

• [ ] Automated daily or weekly querying of major AI platforms for brand-related queries
• [ ] Sentiment scoring system for tracking response quality over time
• [ ] Alert thresholds configured for immediate notification on critical changes
• [ ] Source tracking for all pages cited in AI responses about your brand

Response Resources:

• [ ] Crisis team identified with clear roles and contact information
• [ ] Pre-drafted communication templates for customers, employees, investors, and media
• [ ] Relationship with legal counsel familiar with digital reputation issues
• [ ] Publishing workflow that can produce and deploy corrective content within hours

Prevention Assets:

• [ ] Comprehensive content moat covering all major brand and category queries
• [ ] Consistent entity information across all major platforms and directories
• [ ] Active thought leadership and third-party authority building program
• [ ] Regular AI response audits to catch issues before they become crises

Recovery Playbooks:

• [ ] Documented content amplification strategy ready for activation
• [ ] Third-party publisher relationships that can be leveraged for corrective content
• [ ] Reputation monitoring ChatGPT, Perplexity, Gemini, and Claude tracking dashboards
• [ ] Post-crisis review process to capture learnings and strengthen defenses

Related: E-E-A-T for AI Agents: Establishing Expertise for ChatGPT

Conclusion

The emergence of AI search has introduced a category of brand risk that did not exist two years ago. A single hallucinated claim, an outdated forum post resurfaced by a retrieval algorithm, or a competitor’s strategically positioned comparison page can reshape how thousands of potential customers perceive your brand, and you might not even know it is happening.

But here is the truth that should keep you grounded rather than panicked: AI reputation management follows the same fundamental principles that have governed brand building for decades. Be truthful. Be present. Be authoritative. Be consistent. Be fast when things go wrong.

The tools and tactics are new. The monitoring systems are more complex. The response playbooks require different channels and different timelines. But the core discipline is the same: own your narrative by being the most credible, most comprehensive, and most current source of information about your own brand.

Companies that invest in prevention, that build content moats and maintain entity consistency and proactively shape their AI presence, will weather crises faster and suffer less damage than those who scramble to react after the fact. The best time to build your defenses was six months ago. The second best time is right now.

Do not wait for the 6:47 AM phone call. Build your monitoring systems today. Assemble your crisis team this week. Start auditing what AI models say about your brand before someone else tells you. Because in the age of AI search, your reputation is being shaped in conversations you cannot see, and the only way to influence those conversations is to be prepared before they start.

Related: AI Search vs Google Search: Where SaaS Should Invest in 2026

Ready to Protect Your Brand in AI Search?

Your brand reputation in AI search is either an asset or a liability. There is no neutral ground. WitsCode helps SaaS companies build comprehensive AI reputation monitoring and crisis management systems that detect threats early, respond fast, and build lasting resilience. Book a free AI reputation audit and find out exactly what ChatGPT, Perplexity, and Gemini are telling your prospects about you right now.

FAQ

1. How do I find out what AI models are currently saying about my brand?

The most direct method is to query them yourself. Ask ChatGPT, Perplexity, Gemini, and Claude direct questions about your brand, including comparison queries, reputation queries, and category queries. Do this across different prompt phrasings because AI responses can vary significantly depending on how the question is worded. For ongoing monitoring, build automated systems that run these queries on a schedule and track changes in sentiment and cited sources over time. There is no single dashboard that aggregates this yet, so a combination of manual querying, API-based automation, and traditional brand monitoring tools is the current best practice for comprehensive reputation monitoring ChatGPT and other platforms.

2. Can I contact AI companies directly to correct false information about my brand?

You can, and in some cases you should, but manage your expectations carefully. Most major AI companies have feedback mechanisms, but they do not operate like search engine webmaster tools where you can request specific content changes. OpenAI, Google, and Anthropic all have processes for reporting factual inaccuracies, but these are typically designed for systematic issues rather than individual brand complaints. The most effective approach is indirect: fix the underlying source content that AI models are drawing from, publish authoritative corrective content on your own domain, and ensure your positive content is structured for AI retrieval. Platforms that use real-time retrieval, like Perplexity, will reflect source changes more quickly than models that rely primarily on training data.

3. How long does it typically take for AI models to pick up corrected information?

The timeline varies significantly by platform and by the nature of the correction. Retrieval-augmented generation systems like Perplexity can reflect changes in source content within days to weeks, since they pull from live web results. Large language models that rely on training data, such as base ChatGPT or Claude without web search enabled, may not reflect corrections until their next training data update, which can take months. The practical approach is to optimize for the fastest channels first while maintaining patience with slower ones. Publishing corrective content with strong SEO signals, schema markup, and backlinks from authoritative sources accelerates pickup across all platforms.

4. What should I do if a competitor is deliberately publishing content to damage my AI reputation?

First, document everything meticulously. Capture screenshots, archive pages, and record the timeline of events. Second, assess whether the content is factually false or merely unfavorable. Unfavorable-but-accurate competitor comparisons are a normal part of business competition. False claims, however, may constitute defamation or unfair business practices. For false content, consider legal options while simultaneously executing your content amplification strategy. Publish authoritative responses on your own domain, secure third-party validation from independent sources, and build a body of positive, factual content that overwhelms the negative signal. In cases of sustained deliberate manipulation, consulting with a digital reputation attorney who understands AI systems is advisable.

5. Is AI reputation management something I should handle in-house or hire a specialist for?

The answer depends on the severity and your internal capabilities. Day-to-day monitoring and prevention can and should be managed in-house because no external partner will understand your brand, product, and customers as well as your own team. However, active crises and complex negative citation management situations often benefit from specialist expertise, particularly agencies or consultants who have experience navigating AI platform dynamics and have established relationships with publishers and platforms. The ideal setup is an in-house team managing ongoing monitoring and prevention, with a specialist partner on retainer who can be activated within hours when a crisis is detected. This gives you both the institutional knowledge of an internal team and the specialized skills of an experienced crisis manager.