Modern Search Metrics Glossary: AEO, GEO and AI Scores | Adviora

Glossary of Modern Search Metrics: AEO Score, GEO Score, AI Visibility Index, Share of Model

Search reporting has acquired a lot of new vocabulary in a short time, and not all of it is equally solid. Some of these metrics are published standards with documented thresholds. Some are first-party numbers reported directly by Google. Some are observed or sampled, and depend entirely on your method. And some are composites invented by tool vendors, including us, which are useful for tracking your own direction but meaningless as cross-vendor comparisons.

Confusing the three is how reporting loses credibility. A CMO who is told that a vendor score of 78 is “industry average” will eventually discover there is no industry average, and every other number in the deck gets discounted with it.

This modern search metrics glossary defines 17 terms in plain language, states how each is calculated or observed, says what a good or bad reading looks like, and labels each one as a standard, an observation or a vendor composite. Adviora's GEO & AEO Visibility module reports many of them; we have tried to be exact about which ones are ours.

Why a Modern Search Metrics Glossary Matters in 2026

Because the reporting stack now mixes three kinds of number, and only one of them travels. Label every metric in your deck with its class before you present it.

• Standards — published definitions and thresholds, identical for everyone. Core Web Vitals are the clearest example.
• First-party reported — measured and reported by the platform itself, so accurate but limited to that platform. Search Console's four metrics are the case in point.
• Observed or sampled — you have to go and collect them, and the result depends entirely on your sampling method. Citation coverage and share of model live here.
• Vendor composites — proprietary formulas that summarise many inputs into one score. Useful internally, misleading when compared across tools.

Adviora reports metrics from all four classes, and the honest position is to say which is which. That distinction is also what makes a glossary page worth citing rather than just worth ranking.

AEO Score

AEO Score is a vendor composite that grades how extractable a page's answers are — whether questions are answered in place, in the shape the query implies, and in passages that stand alone.

• How it is observed — scored from on-page structure: question-matched headings, distance between question and answer, passage self-containment and FAQ schema readiness.
• Good or bad — a rising score means your passages are more liftable. It never means Google has selected them, because Google states site owners cannot influence featured snippet selection.

GEO Score

GEO Score is a vendor composite estimating how likely your content is to be retrieved and cited by generative engines when they compose an answer.

• How it is observed — combines passage clarity, factual specificity, crawler accessibility and observed citations across a sampled prompt set.
• Good or bad — a high GEO score with zero observed citations almost always indicates a crawler access problem rather than a content problem. Check robots.txt before rewriting anything.

AI Visibility Index

AI Visibility Index is a vendor composite that rolls your presence across several AI answer platforms into a single tracked number.

• How it is observed — aggregates citation coverage across a fixed prompt set and a fixed platform list, usually weighted by platform importance to your audience.
• Good or bad — only meaningful if the prompt set, platform list and weighting are unchanged between readings. Changing the prompt set resets the series.

Share of Model

Share of Model is your brand's share of all brand mentions inside AI-generated answers, measured across a defined prompt set against a named competitor set.

• How it is calculated — your mentions divided by total mentions of all tracked brands across the same prompts in the same period.
• Good or bad — it is the AI-answer analogue of share of voice, so read it as a brand metric. Rising share against a stable competitor set is the signal; the absolute percentage means little on its own.

Citation Coverage

Citation coverage is the share of sampled prompts in which your domain appears as a cited source in the generated answer, recorded per platform.

• How it is observed — run a fixed prompt set on each platform, record cited or not cited, then divide cited prompts by total prompts.
• Good or bad — assistant answers are non-deterministic, so one reading proves nothing. A three-month trend on identical prompts is evidence; a single screenshot is not.

Citation Sentiment

Citation sentiment is the tone of the text surrounding a mention of your brand in an AI answer — supportive, neutral or cautionary.

• How it is observed — classify the sentence containing the mention and store the verbatim quote alongside the label so the classification can be audited later.
• Good or bad — neutral mentions inside comparisons are normal and healthy. Recurring cautionary language is a trust or review problem that content optimisation will not fix.

Impressions

An impression is counted when a link to your site appears in a search result for a user. It is a first-party metric reported by Google Search Console, not an estimate.

• How it is calculated — counted by Search Console and sliceable by query, page, country, device, search appearance and date.
• Good or bad — impressions rising while clicks fall is the classic signature of answers being resolved on the results page rather than a ranking problem.

Click-Through Rate (CTR)

Click-through rate is clicks divided by impressions for the same query, page or period. In search reporting it is a first-party Search Console metric.

• How it is calculated — clicks ÷ impressions, expressed as a percentage, within whatever segment you have filtered to.
• Good or bad — only compare like with like. Branded queries carry far higher CTR than informational ones, so a blended site-wide CTR is close to meaningless.

Average Position

Average position is the average position of the topmost result from your site, across impressions in the selected period. It is a Search Console metric with a precise definition.

• How it is calculated — Search Console averages the highest-ranking result your site had for each impression, so a page with several ranking URLs is represented by its best one.
• Good or bad — it indicates eligibility, not outcome. Position improving while clicks and CTR fall is a warning, and it is the most commonly misread number in organic reporting.

Core Web Vitals: LCP, INP and CLS

Core Web Vitals are three published page-experience standards: Largest Contentful Paint for loading, Interaction to Next Paint for interactivity, and Cumulative Layout Shift for visual stability. INP replaced First Input Delay as a Core Web Vital in 2024.

• How it is measured — at the 75th percentile of page loads, segmented across mobile and desktop. A single fast lab test tells you nothing about the 75th percentile.
• Good thresholds — LCP at or under 2.5 seconds, INP at or under 200 milliseconds, CLS at or under 0.1. These are published thresholds, so they are directly comparable between sites.

Domain Rank

Domain Rank is a vendor composite, usually on a 0–100 scale, estimating a domain's backlink authority relative to other domains in that vendor's index.

• How it is calculated — derived from the size and quality of the referring-domain graph on a logarithmic scale, so each additional point is harder to earn than the last.
• Good or bad — useful for comparing link prospects inside one tool on one day. It is not a Google metric and no Google ranking system uses it.

Referring Domains

Referring domains is the count of unique domains linking to your site, as distinct from the total number of backlinks, which can be inflated by a single site.

• How it is counted — one domain counts once however many links it sends. Referring IPs are a second diversity check that exposes link networks hosted together.
• Good or bad — steady growth in unique referring domains is a healthier signal than growth in raw backlink volume. Backlink Health 101 covers the full profile.

DoFollow Ratio

DoFollow ratio is the share of backlinks that pass ranking signals, as opposed to those marked rel=nofollow, rel=sponsored or rel=ugc.

• How it is calculated — dofollow links divided by total links in the profile, at either link or referring-domain level; state which you used.
• Good or bad — there is no universally correct ratio. Google requires paid links to be marked nofollow or sponsored, so a healthy profile always contains some.

Spam Score

Spam score is a vendor composite estimating how likely a linking domain is to be manipulative or low quality. Each tool models it differently and the models are proprietary.

• How it is calculated — modelled from patterns the vendor associates with spam; two tools will disagree about the same domain.
• Good or bad — use it as a triage filter for manual review, never as an automatic disavow trigger. Google defines link spam as links created primarily to manipulate rankings.

E-E-A-T

E-E-A-T stands for Experience, Expertise, Authoritativeness and Trustworthiness. It is Google's quality framework, not a score Google publishes. Google states that trust is most important, and that content does not necessarily have to demonstrate all of them.

• How it is observed — through proxies you control: named authors with credentials, transparent ownership, sourcing, and Google's Who, How and Why disclosures, including disclosure of AI use.
• Good or bad — the bar is higher for YMYL topics, meaning content that could significantly affect health, finances, safety or societal welfare. E-E-A-T Decoded breaks this into checks.

Trust Deficit Score (TDS and ETDS)

Trust Deficit Score is Adviora's proprietary measure of the gap between the trust a brand currently demonstrates and the trust its category demands. It decomposes into Effective, Latent and Observed TDS.

• How it is calculated — derived in the CKB Engine from E-E-A-T evidence, technical governance signals and category benchmarks drawn from AGBS.
• Good or bad — it is a diagnostic that selects which lever to pull first, not an industry benchmark. It cannot be compared with any other vendor's trust metric.

Information Gain

Information gain is the amount of genuinely new information a page adds beyond what already exists on the topic. It is a concept rather than a standardised metric.

• How it is observed — compare a draft against the strongest existing sources and count the claims, data points, methods or perspectives that appear nowhere else.
• Good or bad — low information gain is the defining property of mass-produced content, which Google's spam policies treat as scaled content abuse when pages exist primarily to manipulate rankings.

Best Practices for Using These Metrics

1. Label every metric in your reporting with its class: standard, first-party, sampled or vendor composite.
2. Never present a vendor composite as an industry benchmark, and never compare one across tools.
3. Fix the prompt set, competitor set and cadence before your first AI visibility reading.
4. Read impressions, clicks, CTR and average position together; any one of them alone can mislead.
5. Quote Core Web Vitals at the 75th percentile, split by mobile and desktop, or the number is not comparable.
6. Keep one metric per decision. If no decision changes when the number moves, stop reporting it.

The Future of Search Measurement

The gap in this vocabulary is obvious: there is no first-party reporting for AI answer surfaces, so every citation metric in use today is a sample rather than a count. That will either be closed by platforms publishing data, or it will harden into an agreed sampling methodology the way audience measurement did in other media. Until then, the most defensible position is transparency about method — publish your prompt set, your competitor set and your cadence alongside the number.

Primary sources cited:

Google Search Central - AI features in Google Search 

Google Search Central - Featured snippets

Google Search Central - Google common crawlers 

OpenAI - Bots and crawlers