GEO visibility audit: how to map competitors, narrative drift, and quick wins (Weeks 1-2)
If Week 1 is ‘measure before you move,’ Week 2 is ‘stop guessing where you are losing.’
A real GEO visibility audit is not rank tracking with a new label. It is a diagnostic that answers three questions:
- When buyers ask AI for help in your category, do you show up?
- If you do not show up, who shows up instead (and why)?
- Is the way you are described accurate – or is the model freelancing your positioning?
See Also: GEO Content: Intent Pages for AI Mentions and Citation
What a GEO visibility audit includes (the short version)?
A solid audit looks at visibility across three layers:
- Prompt reality: what the AI answers for your money questions.
- Source reality: which pages (yours or competitors) are being used as evidence.
- Site reality: whether your website and brand footprint are making retrieval and citation easy or hard.
Step-by-step: how to run the audit
Step 1: choose your ‘priority surfaces’
Do not audit everything. Audit where your buyers actually search and ask questions.
- Google AI results (if your category is heavily search-driven).
- One conversational assistant your audience uses (often ChatGPT-style).
- One answer engine that cites sources aggressively (often Perplexity-style).
Pick 2-3 surfaces and track them consistently. Consistency beats novelty.
Step 2: run the highest intent prompts first
Start with the prompts closest to purchase:
- best [category] for [use case]
- [brand] vs [competitor]
- alternatives to [competitor]
- how to choose [category]
- [category] pricing / cost
Then expand into implementation and troubleshooting prompts once you understand the battlefield.
Step 3: record four audit signals (the only ones that matter)
| Signal | What you record | Why it matters |
| Mention | Are you named? Where (top, middle, footnote)? | Name placement correlates with buyer recall. |
| Citation/link | Is your site cited? Which URL? | Citations reveal what the model trusts as evidence. |
| Share of voice | Which competitors show up instead? How often? | This becomes your competitive map. |
| Narrative accuracy | How are you described? Any wrong claims? | Narrative drift creates churn and sales friction. |
Step 4: build the competitor map (and stop pretending you know it)
Your true competitors in AI answers are not always the same as your SERP competitors.
Build a map by intent bucket:
- Best pages: who is recommended as ‘best’ and for which use cases.
- Alternatives pages: who is named as an alternative to the category leader.
- Versus pages: which head-to-head matchups show up repeatedly.
- How-to-choose pages: who owns the selection criteria narrative.
- Pricing pages: who sets expectations around cost and ROI.
This map tells you where to attack first. It also tells you which pages you need to publish to win the comparisons.
See Also: GEO Baselining: Week 1 Deliverables That Prove Results
Step 5: audit narrative drift (the silent conversion killer)
Narrative drift is when the model describes you incorrectly.
Common examples:
- Wrong category: you are labeled as a different type of tool or service.
- Wrong positioning: you are framed as ‘enterprise’ when you are SMB (or vice versa).
- Outdated info: old features, old pricing model, old leadership names.
- Made-up limitations: the model invents a weakness you never claimed.
You do not fix narrative drift with wishful thinking. You fix it by tightening your ‘source of truth’ pages and earning stronger third-party echoes.
Step 6: turn findings into a prioritized backlog
The audit output is not a report. It is a ranked list of actions.
A useful backlog separates:
- Quick wins (copy updates, headings, internal links, obvious missing sections).
- Foundation fixes (indexing issues, thin pages, weak About/Service clarity, broken templates).
- New pages (best/alternatives/vs/how-to-choose/pricing gaps).
- Authority work (PR, reviews, partner pages, directories).
What the client should receive (audit deliverables)?
If you are paying for a visibility audit, these are reasonable deliverables to expect:
- A prompt set (or prompt sample) with recorded outcomes by surface.
- A competitor map by intent bucket (who wins best/alternatives/vs/etc.).
- A narrative drift log (what is wrong, where it appears, why it matters).
- A prioritized action backlog with effort and impact estimates.
- 3-10 ‘do this now’ quick wins that can be implemented immediately.
See Also: GEO Reporting: Scorecards That Prove Progress When AI Shift
A simple scoring rubric (so you can quantify progress)
You do not need a perfect model. You need a consistent one.
Here is a scoring approach that works well in practice:
| Metric | Score 0 | Score 1 | Score 2 |
| Mention | Not mentioned | Mentioned (low prominence) | Mentioned (high prominence) |
| Citation | No citation | Citation to non-target page | Citation to target page |
| Narrative accuracy | Wrong/misleading | Mostly accurate, minor issues | Accurate and favorable |
Score each high-priority prompt monthly. The goal is not perfection. The goal is directional movement and fewer ‘zero’ scores over time.
Related reading
- Main article: Done-for-you GEO services: deliverables, timeline, reporting.
- Next Article: On-site upgrades (Weeks 2-6) – making pages crawlable, quotable, and citable.
About The Author
