The AI Visibility Tracking Checklist for Agencies Onboarding a New Client in 2026

Key takeaways

• AI visibility is now a standard client expectation -- agencies that can't show where a brand stands in ChatGPT, Perplexity, and Google AI Overviews are losing the trust conversation early
• The onboarding window (days 1-30) is where most client relationships are won or lost; establishing a clear AI visibility baseline is one of the fastest ways to demonstrate value
• A solid checklist covers five phases: discovery and access, baseline audit, prompt strategy, tracking setup, and reporting cadence
• Most tools on the market only monitor -- the agencies seeing the best results are pairing monitoring with content gap analysis and optimization
• One industry analysis puts average marketing agency churn at 27% annually, and early-stage uncertainty is a major driver; showing concrete AI visibility data early changes that dynamic

---

The question used to be "where do we rank on Google?" Now clients are asking "do we show up when someone asks ChatGPT to recommend a product like ours?" Those are very different questions, and the second one requires a completely different onboarding workflow.

This checklist is built for agency teams handling new client onboarding in 2026. It covers everything from the initial discovery call through the first 90 days of tracking -- with specific steps, the data you need to collect, and the tools worth using at each stage.

---

Phase 1: Discovery and access (days 1-5)

Before you can track anything, you need to understand what you're tracking and get access to the right systems. This phase is about gathering context, not running reports.

What to collect from the client

• A list of their top 5-10 competitors (the ones they actually lose deals to, not just the ones they mention in their deck)
• Their primary product or service categories, including how customers describe them -- not how the brand describes itself
• Any existing SEO data: Google Search Console access, previous keyword research, content audits
• Brand guidelines, tone of voice docs, and any existing content that has performed well
• Their target customer personas, including the specific questions those personas ask when researching a purchase

That last point matters more than most agencies realize. AI models answer questions. If you don't know what questions your client's customers are asking, you can't know whether the client is showing up in the answers.

Access checklist

• Google Search Console (read access minimum)
• Google Analytics or equivalent
• CMS access or at least a content inventory export
• Any previous GEO or AI visibility reports (even informal ones)
• Social and review platform logins if brand monitoring is in scope

Questions to ask on the discovery call

• "Have you ever manually searched for your product category in ChatGPT or Perplexity? What did you find?"
• "Which competitors do you see mentioned in AI answers that you're not?"
• "Do you know if AI crawlers are hitting your site regularly?"
• "Has your organic traffic changed since Google AI Overviews rolled out?"

The answers to these questions tell you a lot about how AI-aware the client already is -- and calibrate how much education you'll need to do alongside the tracking work.

---

Phase 2: Baseline AI visibility audit (days 5-14)

This is the most important phase. You're establishing the "before" state -- the data point everything else will be measured against. Rushing this is the single biggest mistake agencies make.

Define your prompt set

Pick 20-50 prompts that represent how the client's target customers actually search in AI tools. These should include:

• Category-level prompts ("best [product type] for [use case]")
• Comparison prompts ("X vs Y", "[client brand] vs [competitor]")
• Problem-aware prompts ("how do I solve [specific problem]")
• Brand-direct prompts ("[client brand name] reviews", "is [client brand] good for...")

Don't just use the client's marketing language. Use the language real customers use. If you have access to their support tickets, sales call transcripts, or review sites, mine those for phrasing.

Run the baseline across multiple AI models

Check visibility across at least these surfaces:

For each prompt, record: is the client mentioned? Is a competitor mentioned instead? What sources are cited? What's the sentiment of the mention?

Doing this manually across 50 prompts and 7 models is about 350 data points. That's feasible for a one-time audit but not for ongoing tracking. This is where platform tooling pays for itself.

Promptwatch handles this at scale -- tracking prompts across 10+ AI models simultaneously, with citation data, competitor heatmaps, and page-level attribution. For agencies managing multiple clients, that kind of centralized view is what makes the difference between a service you can actually deliver and one that burns out your team.

What to document in your baseline report

• Overall visibility score per AI model (how often the client appears vs. competitors)
• Top 10 prompts where competitors appear but the client doesn't (these are your gap list)
• Which pages on the client's site are currently being cited (if any)
• Which external sources (Reddit threads, review sites, YouTube videos, third-party listicles) are driving competitor citations
• Whether AI crawlers are actively hitting the client's site

That last point is easy to miss. A client can have great content that AI models simply haven't crawled yet. Knowing that changes your prioritization completely.

---

Phase 3: Prompt strategy and gap analysis (days 10-21)

The baseline audit tells you where you are. The gap analysis tells you what to do about it.

Map the content gaps

For every prompt where a competitor appears and the client doesn't, ask: does the client have content that should answer this? The answers fall into three buckets:

1. The content exists but isn't being cited (crawlability or authority issue)
2. The content exists but doesn't actually answer the prompt well (content quality issue)
3. The content doesn't exist at all (content gap)

Each bucket requires a different fix. Don't conflate them -- a client with 200 blog posts might still have massive content gaps for AI-specific queries, because their content was written for Google keyword matching, not for answering the kinds of conversational questions AI models respond to.

Prioritize by prompt volume and difficulty

Not all prompts are equal. A prompt that gets asked 50,000 times a month by high-intent buyers is worth more than one that gets asked 200 times. When building your priority list, weight by:

• Estimated prompt volume (how often this type of question gets asked in AI tools)
• Commercial intent (is this a research query or a buying query?)
• Competitive difficulty (how entrenched are competitors in this answer?)
• Content feasibility (how quickly could the client create or update content to address this?)

Build a content brief for each priority gap

For the top 10-15 gaps, create a brief that includes:

• The specific prompt(s) to target
• What the current AI answer looks like and what sources it cites
• What the client's content needs to cover to be a credible citation candidate
• Recommended format (FAQ, comparison article, how-to guide, product page update)
• Any brand guidelines or tone requirements

This brief becomes the handoff document to whoever is writing the content -- whether that's your agency team, the client's in-house writers, or an AI content tool.

---

Phase 4: Tracking setup (days 14-30)

Once you have a baseline and a gap list, you need to set up ongoing tracking so you can measure progress. This is where most agencies underinvest.

Tracking setup checklist

• Configure your AI visibility platform with the client's brand, competitors, and prompt set
• Set up AI crawler log monitoring (which AI bots are hitting the site, how often, which pages)
• Connect Google Search Console to correlate AI visibility changes with organic traffic
• Set up alerts for new competitor citations on high-priority prompts
• Configure page-level citation tracking (which specific pages are being cited)
• Set up offsite citation monitoring (Reddit, YouTube, third-party listicles that mention the brand)

The crawler log piece is worth calling out specifically. Most agencies skip it because it feels technical, but it's genuinely useful. If you publish new content and an AI crawler visits the page within 48 hours, that's a signal the content has a real chance of being cited. If the crawler never comes, you have a discoverability problem to fix before the content can work.

Tool options for tracking

There are a lot of tools in this space now. Here's a practical breakdown of what to consider:

The honest answer is that most monitoring-only tools will show you the data but leave you figuring out what to do with it. If your agency is positioning AI visibility as a managed service (not just a report), you need a platform that connects the data to action.

Set up your reporting infrastructure

Before the first client report is due, decide:

• How often will you report? (Weekly snapshots + monthly deep dives is a common cadence)
• What metrics will you lead with? (Visibility score, citation count, prompt coverage, traffic from AI referrals)
• Who on the client side receives the report?
• Will you use white-labeled exports, a live dashboard, or a custom deck?

Getting this agreed upfront prevents the awkward "what does this number mean?" conversation on the first reporting call.

---

Phase 5: First 90-day optimization cycle

Onboarding doesn't end at day 30. The first 90 days is where you prove the model works.

The optimization loop

The most effective agencies run a tight cycle:

1. Identify the highest-priority content gaps from the prompt analysis
2. Create or update content to address those gaps
3. Monitor AI crawler activity to confirm the content is being discovered
4. Track whether citations appear in AI responses for the target prompts
5. Connect citation growth to traffic and revenue where possible
6. Repeat with the next batch of gaps

This loop -- find gaps, create content, track results -- is what separates agencies that can show ROI from agencies that show dashboards.

30-day check-in deliverables

• Updated visibility scores vs. baseline
• First content pieces published and indexed
• Crawler log summary (which AI bots visited, which pages)
• Any early citation wins to highlight
• Revised priority list based on what's moved

60-day check-in deliverables

• Visibility trend lines (are scores moving?)
• Citation count by AI model
• Traffic from AI referrals (Perplexity, ChatGPT, etc. in analytics)
• Competitor comparison: are gaps closing?
• Content performance: which new pages are being cited?

90-day review deliverables

• Full before/after comparison vs. baseline
• ROI narrative: visibility gains tied to traffic and lead data where available
• Roadmap for the next quarter
• Expanded prompt set based on what's working

---

Common mistakes to avoid

A few things that consistently trip up agencies new to this service:

Treating AI visibility like keyword rankings. They're related but different. A page can rank #1 on Google and never appear in a ChatGPT answer. The citation logic in AI models is driven by authority, content quality, and how well a page answers a specific question -- not just domain authority or backlink counts.

Setting up tracking before defining the prompt set. If you track the wrong prompts, you'll get clean data that tells you nothing useful. Spend the time upfront to define prompts that reflect real customer intent.

Ignoring offsite citations. A significant portion of AI citations come from Reddit threads, YouTube videos, industry listicles, and review sites -- not the client's own domain. If competitors are winning because they're well-represented on G2 or a popular subreddit, that's a different fix than updating the client's blog.

Reporting visibility scores without context. A visibility score of 34% means nothing to a client unless they know the competitor is at 61% and the industry average is 28%. Always frame numbers in context.

Skipping the crawler log setup. It feels optional. It isn't. Knowing whether AI bots are actually reading your new content is the difference between waiting and wondering vs. diagnosing and fixing.

---

The deliverable checklist: what to hand a client at the end of onboarding

By day 30, a well-run AI visibility onboarding should produce:

• Baseline visibility report (scores per AI model, per prompt category)
• Competitor heatmap (who's winning which prompts and why)
• Top 15 content gaps with briefs
• Crawler log summary and any technical issues identified
• Offsite citation audit (where competitors are being cited that the client isn't)
• Tracking dashboard configured and shared
• Reporting cadence agreed and first report scheduled
• 90-day roadmap with content priorities and expected milestones

That's a concrete, defensible deliverable set. It shows the client exactly where they stand, what's being done about it, and how you'll measure progress. That's what builds trust in the first 30 days -- and trust is what keeps clients past month six.

The agencies that are winning new business on AI visibility right now aren't the ones with the fanciest pitch decks. They're the ones who can walk into a kickoff call, run a baseline audit within a week, and show the client something real -- a gap list, a competitor comparison, a content plan. That's the whole game.