The new performance problem: AI is your new ad network… and it’s off-script
Search results, social feeds, inboxes, and even operating systems are quietly turning into AI answer engines:
- Google rolling out Gemini into Search and AI Overviews
- ChatGPT, custom GPTs, and “AI Mode” experiences
- Microsoft Copilot wired into Windows, Edge, and Office
At the same time:
- Doctors are finding AI Overviews inventing career-damaging claims
- Ahrefs is showing how brand visibility changes inside AI experiences
- AI misinformation experiments are proving how confidently wrong these systems can be
For performance marketers, this isn’t a thought experiment. It’s a new media channel with three brutal properties:
- You don’t fully control the creative (the AI writes it).
- You don’t fully control the placement (the AI chooses when you appear).
- You don’t fully control the tracking (the AI often sits between you and the user).
This article is about treating AI surfaces like a real performance channel: measured, optimized, and defended – not just “something SEO will figure out.”
The shift: from “rank in search” to “be the answer engine’s default”
Traditional search optimization is about ranking a URL. AI answer optimization is about becoming:
- The entity the model associates with a topic
- The default brand it recommends for a job-to-be-done
- The source it cites when it needs a fact, stat, or example
That’s a different game than “write a better title tag” or “add more keywords.” It’s closer to:
- PR: shaping how you’re described
- Conversion rate optimization: shaping how offers are framed
- Brand safety: preventing harmful or off-brand claims
And it’s already affecting performance:
- AI Overviews can answer the query without a click, cutting search traffic
- ChatGPT and Copilot can recommend tools and vendors directly, bypassing comparison pages
- Custom GPTs and AI workflows can become “middleware” between demand and your site
Step 1: Audit how AI currently talks about you
Before you optimize, you need to see the mess. Treat this like a new channel audit.
Run a structured “AI brand visibility” audit
For your top markets and languages, test:
-
Branded queries
“What is [Brand]?” “Is [Brand] legit?” “Alternatives to [Brand]” “Who is [Founder]?” -
Category queries
“Best [category] tools for [segment]” “Top [category] platforms under $X” -
Use-case queries
“How do I [job-to-be-done]?” “What’s the best way to [outcome your product drives]?”
Run these across:
- Google Search with AI Overviews (where available)
- ChatGPT (default and GPT-4o), including “AI Mode” style prompts
- Microsoft Copilot on desktop and mobile
Score what you see
Create a simple scoring sheet:
- Presence: Are you mentioned at all? (0 = no, 1 = yes)
- Position: Are you a primary recommendation, or buried in a list?
- Framing: How are you described? (accurate / incomplete / wrong / harmful)
- Attribution: Are your owned properties cited? (site, docs, blog, help center)
- Click path: Is there a clear path to your site or product, or are you “summarized away”?
This gives you a baseline “AI channel health” view: visibility, quality, and risk.
Step 2: Fix the dangerous stuff first (brand safety for AI answers)
If AI is inventing harmful claims about you or your people, that’s not an SEO problem. It’s a crisis problem.
Prioritize harmful or materially wrong outputs
Flag and screenshot:
- False legal, medical, financial, or security claims about your company or team
- Made-up scandals, lawsuits, or “reviews”
- Wrong pricing, guarantees, or policies that could create liability
Respond with a structured playbook
For each serious issue:
-
Document
Save timestamps, prompts, full responses, and any cited sources. -
Correct the public record
Publish a clear, factual statement on your site (FAQ, help article, or blog) addressing the specific claim. -
File formal feedback
Use the platform’s feedback/reporting tools with links to your correction page. -
Loop in legal/PR
Especially for defamation, regulated claims, or anything that could affect hiring, fundraising, or customer trust.
This isn’t guaranteed to fix the model quickly, but it:
- Creates authoritative content for future model training
- Gives you a paper trail with vendors and partners
- Protects you if a customer or prospect cites the AI output later
Step 3: Optimize for “AI answer inclusion” like a performance channel
Once the fires are out, treat AI answer surfaces like a hybrid of SEO and affiliate: you want to be the default recommendation for high-intent queries.
Design content for models, not just humans and crawlers
Models like clean, structured, unambiguous information. That’s good news – you can engineer for it.
-
Entity clarity
Make it painfully obvious who you are:- Consistent name, tagline, and category across site, LinkedIn, G2/Capterra, Crunchbase, etc.
- Clear “About” and “Who we’re for” pages with structured data (Organization, Product, Person schemas).
-
Job-to-be-done pages
Create pages that map tightly to how users ask AI for help:- “How to [job] for [segment]”
- “Best way to [outcome] without [pain]”
Spell out:
- Who this is for
- What they’re trying to achieve
- Why your approach is different
-
Comparison and “best of” content
AI answer engines love lists and comparisons. If you’re not in the lists they see, you’re invisible.- Create honest comparison pages (you vs. key alternatives) with clear tradeoffs.
- Publish “best tools for X” content where you appear alongside others, not just as the hero.
-
Evidence and numbers
Models latch onto stats and concrete claims. Feed them:- Case studies with clear numbers (“37% more inquiries”, “22% lower CAC”)
- Methodology sections that explain how you got the numbers
Think like a media buyer: prioritize topics by intent and value
Don’t try to “AI-optimize” everything. Start where the money is.
-
Map high-value queries
Use your paid search and SEO data:- Top converting keywords (by revenue or qualified pipeline)
- High-intent modifiers: “best”, “vs”, “alternative”, “for [segment]”
-
Check AI visibility for those queries
Where you’re missing or misrepresented, that’s your “AI channel gap.” -
Assign owners and timelines
Treat each high-value query cluster like a campaign:- Goal: “Be included as a recommended option for X”
- Inputs: content, PR, structured data, reviews, partnerships
- Review cadence: re-check AI answers monthly
Step 4: Adjust your tracking and attribution expectations
AI answer engines introduce a new kind of “dark social” – call it “dark AI.” People discover you through AI, then show up as:
- Direct traffic
- Brand search
- “Typed” URLs on mobile
- Referral from random AI tools with no UTM support
Instrument for AI-influenced demand
You won’t get perfect tracking, but you can get directional signal.
-
Update “How did you hear about us?”
Add options like:- “Google AI Overview / AI answer”
- “ChatGPT / Copilot / other AI assistant”
Keep it short and visible in high-intent forms and sales discovery.
-
Watch branded search and direct lifts
When you ship major content or PR that should influence AI, watch:- Branded search volume
- Direct traffic to product and pricing pages
- Changes in “unattributed” or “self-reported AI” deals
-
Tag AI-focused content separately
Use content groupings or custom dimensions for pages built primarily to influence AI answers, so you can track their indirect impact.
Step 5: Use AI tactically without automating yourself into a ditch
The other half of this story: you’re not just being summarized by AI – you’re also using it in your own workflows. That’s where performance teams quietly burn money.
Guardrails for AI in campaign production
AI is great for:
- Variant generation (ad copy, hooks, angles)
- Summarizing research and customer calls
- Drafting briefs and testing hypotheses
It’s terrible when you:
- Let it invent product claims or compliance language
- Use it to “optimize” tracking or campaign settings without human review
- Copy-paste outputs into live campaigns without QA (see: “double-checking campaign settings matters”)
Practical guardrails:
- Every AI-generated asset gets a human owner and a checklist (claims, pricing, brand voice, compliance).
- Any AI-suggested change to budgets, bids, or tracking must be reviewed in a change log.
- Custom GPTs you use internally should be trained on your approved messaging and constraints, not the open internet.
Custom GPTs as internal “performance copilots”
Instead of letting general-purpose models hallucinate your strategy:
- Build internal GPTs that:
- Ingest your product docs, brand guidelines, and offer library
- Know your funnel stages, KPIs, and constraints
- Are explicitly banned from inventing numbers or claims
- Use them to:
- Draft experiment ideas with hypotheses and metrics
- Summarize performance reports for stakeholders
- Generate structured test matrices (channels x audiences x offers x creatives)
The goal: AI accelerates your thinking and production, but humans still own the bets and the risk.
What to do this quarter
To turn all of this into an actual plan, here’s a 90-day roadmap.
Weeks 1-2: Visibility and risk audit
- Run the AI brand visibility audit across Google AI Overviews, ChatGPT, and Copilot.
- Document harmful or materially wrong outputs and start the correction process.
- Identify 10-20 high-intent queries where AI answers matter most to your pipeline.
Weeks 3-6: Content and entity cleanup
- Standardize your brand and product entities across your site and major directories.
- Ship or update:
- About / Who we serve
- 3-5 job-to-be-done pages
- Key comparison pages
- Add structured data where it’s missing (Organization, Product, Person).
Weeks 7-10: Measurement and internal AI guardrails
- Update “How did you hear about us?” to include AI options.
- Set up content groupings for AI-oriented pages.
- Define internal rules for AI use in campaign production and tracking.
- Optionally, build a simple internal GPT for briefs and experiment ideation.
Weeks 11-12: Review and iterate like a media channel
- Re-run your AI visibility checks for the high-intent queries you targeted.
- Compare branded search, direct traffic, and self-reported AI influence vs. baseline.
- Decide where AI answer optimization sits in your org (SEO? Growth? Brand?) and give it an owner.
AI isn’t just another tool in your stack. It’s a new distribution layer between your budget and your buyers. Treat it like a channel, not a toy – and make sure, at minimum, it’s not out there lying about you.