Reading your Reffed audit.
A Reffed audit is the automated version of Lesson 3's 8-prompt audit, run across 6 engines simultaneously. This lesson walks through every number on the report so you know exactly what to fix first.
Get a free audit first. The rest of this lesson makes more sense if you have your own report open in another tab.
Run a free audit →The overall score (0-100)
The headline number is a weighted composite. The weights, in order of impact:
- Mention rate across 6 engines (40%). How often your brand appears when AI engines are asked relevant questions about your category.
- Entity signal strength (25%). How consistently AI engines verify who you are across Wikipedia, LinkedIn, Crunchbase, Google Business Profile, and other entity sources.
- Content extractability (20%). How citation-ready your top pages are — schema markup, answer-block structure, statistics, quotations.
- Citation diversity (15%). How many distinct third-party sources mention you — Reddit, G2, industry lists, editorial coverage.
Score interpretation:
| Score | What it means |
|---|---|
| 85-100 | AI search winner. Defend position. Focus on share of model in core categories. |
| 70-84 | Strong. You have entity signal and citations. Plug specific gaps to break into top 3. |
| 50-69 | Mid-pack. Inconsistent citations. Full playbook — entity, content, third-party — applies. |
| 30-49 | Weak. Most prompts return competitors instead of you. Start with entity foundation. |
| 0-29 | Invisible. AI engines do not recognize the entity. Foundational work needed. |
Per-engine breakdown
Below the overall score, Reffed shows your performance on each of 6 engines. This is where the strategy lives. A 65 overall could mean any of these very different situations:
- Even distribution (e.g., 65/65/65/65/65/65). Consistent visibility everywhere. Lift the floor across the board.
- ChatGPT 90, Perplexity 30, others 65. You have brand authority but fact density is low. Add statistics and recent updates.
- Google AI Overviews 95, Claude 20, others 65. Strong schema, weak Wikipedia/expert presence. Pursue editorial quotation and Wikipedia entry.
- Perplexity 90, ChatGPT 30, others 65. Fact-rich content but weak brand validation. Pursue G2, Reddit, and third-party mentions.
The pattern matters more than the average. Read the per-engine breakdown to identify which signal is weakest, then attack that signal specifically.
The four subscores
Reffed breaks the overall score into four components matching the weights above:
Mention rate
Percentage of times your brand is named when relevant prompts are run. This is the most volatile subscore week-to-week. Movement here usually reflects training data updates or new third-party content surfacing.
Entity signal
Composite of: presence on Wikipedia, LinkedIn company page, Crunchbase, Google Business Profile, Apple Business Connect, and Bing Places. Plus consistency — same business name, same founding year, same description, same location across all sources.
This is the cheapest subscore to fix and the highest-leverage. A 20-point entity signal lift typically lifts the overall score by 10-15 points within 60 days.
Content extractability
Audit of your top 10 pages for: question-formatted H2 headings, answer blocks of 120-180 words immediately following each H2, schema markup (FAQPage, Article, HowTo, Organization), statistics with attribution, direct quotations from named experts.
The Princeton GEO research found this pattern alone drives up to 40% citation visibility lift without changing word count.
Citation diversity
Count of distinct third-party sources mentioning your brand: Reddit threads, G2/Capterra/Clutch reviews, industry list inclusions, editorial coverage, Wikipedia. The 2.8× citation multiplier from Princeton applies here — being mentioned across 4+ platforms increases ChatGPT citation likelihood by 2.8×.
The mention check section
Below the subscores, Reffed shows the actual AI responses it received for each prompt, on each engine. This is the most useful part of the report for execution.
Look for three patterns:
- Misrepresentation. AI knows you but gets a fact wrong (year, location, product focus). Fix the source of the bad fact — usually Wikipedia, LinkedIn, or your homepage.
- Competitor substitution. AI recommends a similar competitor instead of you. Look at why — is the competitor cited in more "alternatives to X" articles? More Reddit threads? More G2 reviews?
- Total absence. AI says "I do not have specific information about this brand." Entity foundation work needed.
Turning the audit into an action plan
A working pattern that we have seen produce results across audits:
Week 1-2: Fix entity consistency. Sync your name, founding year, headquarters, and description across Wikipedia, LinkedIn, Crunchbase, Google Business Profile. Cost: ~6 hours of work.
Week 3-4: Restructure your top 5 pages. Convert generic H2s to question-formatted H2s. Add 120-180 word answer blocks. Add 1-2 statistics with source attribution to each section. Add Article + FAQPage schema. Cost: ~12 hours of work.
Month 2-3: Pursue third-party presence. One Reddit thread answering questions in your category per week. One G2/Capterra review per existing customer. One industry list inclusion per month (pitch editors directly). Cost: ~4 hours per week, ongoing.
Month 3-6: Re-audit monthly. Watch which subscores move. Adjust focus to the slowest-moving subscore.
Free audits are point-in-time. Reffed Watch ($29/month) re-runs your audit weekly across all 6 engines, tracks subscore movement, and alerts you when a competitor overtakes you on a key prompt.
What you can do this week
Run a free Reffed audit. Note your overall score, your per-engine pattern, and your lowest subscore. Lesson 5 turns these numbers into a specific 30-day plan with weekly checkpoints.