Read and defend an AGS score

Step 1 : Find the AGS score on your dashboard

Open any completed audit. At the top of the results page, you see the global AGS score (0–100) along with its Hero Grade letter (A+ to F). The number itself is the geometric mean of three composite dimensions: Presence, Impression, Quality. The Hero Grade is its qualitative label, designed to be readable in one glance by a non-technical reader.

Step 2 : Read the 3 dimensions P / I / Q

Just below the global score, three sub-scores appear:

• P — Presence : is your brand mentioned? At what position? In how many words? Computed across all queries (even those where the brand was not mentioned).
• I — Impression : when mentioned, is the information informative, unique, relevant? Computed only on queries where the brand was mentioned.
• Q — Quality : is the sentiment positive? Are the cited facts accurate? Computed only on mention queries.

Step 3 : Check the GRC inter-judge reliability coefficient

Each AGS score is published with a GRC (Grade Reliability Coefficient) between 0 and 1. It measures the agreement between the five AI judges that scored the same items. Reading guide:

• GRC ≥ 0.80 : excellent agreement, score directly defensible.
• 0.60 ≤ GRC < 0.80 : good agreement, normal for noisy domains.
• GRC < 0.60 : weak agreement — check if your audit has too few prompts (less than 20) or if the brand is in a controversial domain.

When you defend your number in front of a client, citing the GRC is exactly what separates a serious audit from a magic black-box number. No other GEO/AEO platform we know of publishes this coefficient.

Step 4 : Read ASR (anti-drift correction)

Right next to the AGS score, you see ASR = AGS minus the model drift measured during the same week on five stable anchor brands. When OpenAI updates GPT-4o or Google adjusts Gemini, every brand’s score moves — even those that did nothing. The anchor set isolates this drift, and ASR shows you the score corrected for it.

Concrete reading : if your AGS climbed from 62 to 71 between two audits but ASR climbed from 62 to 67, only 5 points of progression are yours. The remaining 4 points come from the models becoming more generous on average that week.

Step 5 : Verify the SHA-256 configuration hash

At the bottom of the audit detail page, in the « technical signature » section, you find the judge_config_hash: an SHA-256 fingerprint of the judges pool, weights, and rubric files used during this audit.

• Same hash between two audits → scores directly comparable.
• Different hash → the formula or judges have changed; the dashboard explicitly logs it and offers to rebaseline if needed.

When a competitor tells your client « your score went from 62 to 71 », they have no way to know whether the brand actually progressed or the formula changed in between. With AGS, this is a non-issue: the hash is the proof.

Step 6 : Defend your numbers in front of a client

When you present an audit to a marketing director or a steering committee, your slide should always include four numbers, in this order:

1. AGS + Hero Grade : the global number, easy to memorise.
2. P / I / Q : the breakdown, to know where to act.
3. GRC : the methodological credibility signal.
4. ASR : the score corrected for AI drift, for a fair temporal comparison.

Anyone questioning your number gets an answer rooted in code, not marketing : you can show the GitHub repo of the AQA standard, the methodology white-paper, the SHA-256 hash on the audit, and the published GRC. That’s how we replace « trust me » with « verify it ».

Going further

• Article: AGS in depth (GRC, anchor set, SHA-256)
• Multilingual audit: 4 European markets in 1 audit
• Glossary: Hero Grade (A+ to F)
• Glossary: Native Score
• AQA Hub: open standard for AI-readable FAQs