ARO Index™
ARO INDEX METHODOLOGY

The ARO Score Method (v3 · July 2026)

What the ARO Score measures, how audits are structured, and why the method changed in 2026.

Published · Updated

What we measure

Selection, not mentions.

When an AI is asked to recommend a business in a category and city, does it choose this one. Being named is visibility. Being chosen is selection.

Two call types, one clear line

Evaluation calls and cold-selection calls answer different questions.

Evaluation calls (the ARO Score below) include the business's own site content by design and ask a model to judge whether it would select that business for a real buyer query. Because the model is working from real site content instead of guessing, every audited business starts from a score floor rather than a blank page. This measures whether a business is recommendable.

Cold-selection calls (used in ARO Index research reports) fire the exact same locked queries with no site content supplied and no business name planted in the prompt, and only count businesses the model volunteers entirely on its own. This measures whether a business is actually recommended. See the Observed-Selection Census method below for the full breakdown.

How an audit runs

12 independent reads per audit.

Each audit runs 3 separate queries per model across 4 models (ChatGPT, Claude, Gemini, Perplexity), each including the business's own site content and live web retrieval, producing 12 independent reads. Queries are never bundled, so models cannot self-anchor and manufacture false agreement.

Model Access Surface

What each model is actually working from.

Model API model string Retrieval state Confirmed date
ChatGPT gpt-5.6-luna Live web retrieval (web_search tool) 22 Sep 2026
Claude claude-sonnet-5 Live web retrieval (web_search tool) 22 Sep 2026
Gemini gemini-3.8-flash Live web retrieval (google_search tool) 22 Sep 2026
Perplexity perplexity/sonar (Agent API) Search-grounded at provider level 22 Sep 2026

Confirmed by direct audit of the audit runner source code.

Panel change log

Sept 14, 2026: Perplexity moved from sonar-pro (Chat Completions) to perplexity/sonar (Agent API) mid-wave. sonar-pro sunsets Sept 27, 2026. No Vol 3 row ran on the new string.

Sept 16, 2026: live pipeline moved from the July panel to the Sept panel above by deliberate operator decision. Grades against current consumer base models.

Sept 22, 2026: the Gemini leg ran gemini-3.8-flash, Google's current GA API model. On Sept 22, 2026, the free Gemini consumer app defaulted to 3.6 Flash on two separate accounts. Whether 3.6 Flash was the consumer default on Sept 14 is not confirmed.

October 18, 2026: Scheduled: the November 1 to 3, 2026 census wave moves ChatGPT from gpt-5.6-luna to gpt-6-luna, matching the free consumer default since October 8, 2026. All eight markets run on the v2.0 category appendix (164 categories). Atlanta, Denver, Sacramento and Miami skip one monthly cycle (last run September 18, 2026 on the 82-category appendix) so the whole Index moves on one boundary. Results across this boundary are not directly comparable.

November 2, 2026 (amends the October 18, 2026 entry): The census_20261102 wave, planned for November 3 to 5, 2026, moves three of the four query models to each product's current free-tier default. ChatGPT moves from gpt-5.6-luna to gpt-6-luna, the free ChatGPT default since October 8, 2026. Claude moves from claude-sonnet-5 to claude-sonnet-5-5, the model on Claude's free plan. Gemini moves from gemini-3.8-flash to gemini-3.5-flash-lite, since Google lists Flash-Lite as the free Gemini model. Perplexity stays on sonar, the closest API equivalent to free Perplexity; it is not an exact match. Twelve markets run on the v2.0 category appendix (164 categories): Aiken, Atlanta, Charleston, Denver, Fresno, Greenville, Manhattan, Miami, Nashville, Parker, Sacramento and Sarasota. Newton is excluded because it has no categories yet. Results across this boundary are not directly comparable.

November 10, 2026: census_20261102 (November 2026): query models ChatGPT gpt-6-luna, Claude claude-sonnet-5-5, Gemini gemini-3.5-flash-lite, Perplexity sonar. Sonar is the closest API equivalent to free Perplexity, not an exact match. Every answer is read by the same extraction model, claude-haiku-5-5, which only extracts and matches business names and never answers a query. This is a panel change, not a methodology change.

All four models run with live web search enabled on every audit call, using the business's own site content alongside what each model retrieves on its own. The ARO Score reflects that combined read, not training data alone.

v3 grounded became the default method for every new audit on July 31, 2026. Audits completed before that date remain on v2 and are labeled accordingly. v2 and v3 scores are not directly comparable, since v3 adds live web retrieval that v2 did not have. Businesses move to v3 automatically as their next audit completes.

Gemini's reasoning effort is fixed to a low, explicit setting on every audit call. Left unset, the model defaults to its highest reasoning tier, which increases latency and output token volume without changing the scoring rubric it is asked to apply.

How the score is built

Majority across queries, not a single read.

Per model, a 2-of-3 majority across the three queries decides whether that model recommends. The score reflects selection across the full range of queries, not a single lucky one.

Why three queries

One phrasing is one camera angle.

Models are sensitive to how a question is asked. Three separate queries show how a business is interpreted across how buyers actually search.

Query phrasing

Three intent angles, not three synonyms.

Each locked query bank carries 3 phrasings built on the same three buyer-intent angles: a direct query ("best [category] in [city]"), a referral query, and a seeking query ("looking for a [category] in [city]"). This is the same 3-phrasing structure used for both evaluation calls and cold-selection calls, so the two tracks above are always asking about the same underlying buyer intents.

The referral phrasing varies by category type. Place categories (restaurants, hotels, tours, coffee shops, and similar) use "can you recommend a [category] in [city]." Provider categories (professionals and service providers people hire directly) use "who would you recommend for [category] in [city]." Both ask the same thing in the way a real buyer would actually phrase it for that kind of business.

Why we run multiple queries

Not all models respond the same way twice.

A June 2026 variance study across five markets and four models found Gemini and Perplexity stable enough for a single read -- their top picks held consistent across repeated queries on the same business. ChatGPT and Claude were less stable, especially in crowded categories where many businesses have similar web presence.

The response: 3 separate queries per model, with a 2-of-3 majority required for a recommendation to count. One lucky answer does not move your score.

ARO Score™ formula

Four factors, published weights.

ARO Score™ is the diagnostic composite inside the ARO Report. It is 0 to 100 and combines four factors:

  • Recommendation rate (40%)
  • Position strength (20%)
  • Confidence (20%)
  • Signal quality (20%)

Recommendation rate carries the most weight because being chosen is the point. The other three factors reflect how prominently, how consistently, and how reliably that selection holds. ARO Score™ explains what may be driving a business's ARO Index Score. It is not the public benchmark itself.

ARO Index Score formula

The public benchmark, rebuilt on observed outcomes.

The ARO Index Score is the public benchmark used to rank and compare businesses within a market and category. As of July 2026, it is calculated entirely from observational inputs, not from the ARO Score™ diagnostic composite. Individual reads come from the audit method current at the time of each audit (v2 before July 31, 2026, v3 grounded after). The ranking formula below changed in July 2026 and is applied to both.

The ARO Index Score is 0 to 100 and combines four factors:

  • Recommendation rate (45%)
  • Cross-model consistency (20%)
  • Recommendation position (25%)
  • Recency (10%)

Recommendation rate carries the most weight because being recommended is the point. Cross-model consistency and recommendation position reflect how reliably and how prominently that recommendation holds across ChatGPT, Claude, Gemini, and Perplexity. Recency weights recent audits more heavily than older ones.

Position strength

Where you appear in the list matters.

A recommended business named first earns more weight than one named fifth. Higher ranks earn descending credit down the list. A business that is recommended but given no clear rank receives a neutral middle weight. This rule is fixed and applied identically to every business.

Signal quality

Four grouped factors, not a checklist.

Signal quality captures the site-side factors that affect how well AI platforms can read and trust a business. Four groups:

  • Content clarity
  • Technical crawlability
  • Structured data
  • Authority signals

No individual dimension score is published. The aggregate drives the component weight.

Cadence

Monthly by market.

Each market is re-audited monthly. Rankings update when new audits complete.

The query bank is versioned separately from the model panel; a bank version change and a model panel change are each a comparability boundary, and the Index moves both on the same wave where possible.

Conflict-of-interest rule

No business can pay to change its score.

Weights are fixed and applied identically to every business, client or not. Buying a subscription changes what advice you receive; it does not touch the audit engine or change how your score is calculated. Measurement and services are kept separate by design.

Data currency

A score reads a moment.

Scores are dated and re-run. Models update, competition shifts, questions evolve. A score reads a moment; it is not a permanent grade.

The original single-query method (v1, through early 2026) is preserved below for reference.

Used in published research (e.g. Charleston 2026)

A separate, stricter measurement than the ARO Score.

The ARO Score above audits a business against its own category queries. The Observed-Selection Census, used in ARO Index research reports, asks a different question: when a real buyer types an unprompted query with no business name planted in it, which businesses does the model name on its own. Every published research number (Charleston 2026 and future city reports) traces back to this method, not the ARO Score formula above.

Data correction under current scoring (October 2026)

What the direct audit measures.

Data correction under current scoring, October 3, 2026. Earlier versions of this page described the direct-audit lens as running a business's category queries and recording whether the models name it. In fact the models were shown the business's website content and schema alongside the queries and asked whether they would recommend it. The measurement and every figure (688 businesses, 76.2%, 23.8%, 12.9%, 87.1%) are unchanged; only the description is corrected, and this is not a new scoring version. The direct lens measures whether a business is recommendable from its own page. The cold lens measures whether a model volunteers it with no page shown.

Correction note, October 2, 2026

Query banks and the three intent variants.

The rule. Every ARO audit asks ChatGPT, Claude, Gemini and Perplexity about one type of business in three ways: a direct question ("best dentist in Charleston"), a referral question ("who would you recommend for a dentist in Charleston"), and a seeking question ("looking for a dentist in Charleston"). A model counts as recommending a business when it says yes on at least 2 of those 3 questions. The ARO Score reflects how many of the 4 models do.

What we found. On October 1, 2026 we audited all 572 query banks behind the ARO Index. Two problems showed up:

  • 125 banks, in 14 groups, asked about different business types instead of one type in three ways. The Fitness & Wellness bank, for example, asked about "best gym", "personal trainer" and "yoga studio", so a yoga studio was graded partly on gym and trainer questions it would never be the right answer to. Med spa, law firm, marketing agency and several other banks had the same flaw.
  • 308 of the 572 banks did not use three distinct question styles (direct, referral and seeking), so many audits did not deliver the three variants this page describes.
  • A smaller set of banks ran questions for a different business type than their name, and 18 banks had a city written into the question text.

What changed.

  • October 1, 2026: 148 banks were rebuilt so that each asks about a single business type in the three styles above, with the city filled in per market (144 rebuilt plus 4 new Fitness & Wellness banks: Gym, Yoga Studio, Pilates Studio and Fitness Studio). 695 old question rows were retired, 20 groups of near-duplicate banks now point to one existing broader bank, and 3 banks with no audits behind them were retired outright. 11 further banks (Cleaning Services, Criminal Defense, Deep Tissue Massage, DUI, Electrical, HVAC, Immigration, Landscaping, Personal Injury, Plumbing, Roofing) already asked one type in three styles and were kept as they were.
  • October 2, 2026: 77 more single-type banks were built (49 on the first pass, 28 more later the same day) from the same template, including Med Spa, Plastic Surgery, Law Firm, Real Estate, Mental Health and Women's Health, which were still mixed-type banks on October 1. 50 of those replaced older banks (159 old question rows retired); 27 are new types that had no bank. The question rows for the 11 banks above were stamped as meeting the template without changing their wording.
  • The Aiken, SC census is the first run on the new one-bank-per-row panel (Panel Protocol v2.0, 164 rows). The earlier census series (July and September 2026, 82 rows) used the older multi-type banks and is closed; it is not trended against the new series.

What is not finished. All 164 census banks now use the template. 290 other banks across the wider index have not been verified against the template; audits that used them are labeled with the exact question that was asked.

Historical reads are kept. No earlier audit result was deleted or relabeled. Every stored read still shows the exact question that was asked, and those reads remain available as the "before" record. For example, a stored read for "best gym Greenville SC" stays as a valid direct-question read.

Re-audits. About 516 latest audits were scored against an affected bank: 81 in Fitness & Wellness and 435 in other categories. Their scores will be re-run under the current panel and the current scoring method. Fitness & Wellness is first; the others follow as banks are rebuilt. Until a business is re-audited, its earlier score stays visible with its original questions, and it may move up or down when re-run.

This is a data correction, not a new methodology. The scoring method, the 4-model panel (ChatGPT, Claude, Gemini, Perplexity) and the 2-of-3 rule are unchanged. We corrected the questions the existing method was supposed to be asking, and the census moved from parent categories to single business types, disclosed under Panel Protocol v2.0.

Last updated: October 2, 2026.

Two lenses, not one number

Direct audit vs. cold market query.

Direct audit: show each model the business's own website and its category queries, and ask whether an AI assistant would recommend it. A model counts when it says yes on at least 2 of 3 queries. In the Charleston census, 76.2% of the 688 audited businesses are recommended by at least one model this way. This measures whether a business is recommendable from its own page; the model is shown the business rather than finding it.

Cold market query: run the unprompted "best [category] in [city]" a buyer would actually type, with no business name planted in the prompt, and record which businesses the models name on their own. In the Charleston census, only 12.9% are ever selected this way. No business name, page text or schema is in the prompt. The difference between the two is the page evidence the model is shown.

A business can pass the direct-audit lens and still fail the cold-query lens. That gap is the finding: recommendable is not recommended. Reports lead with the cold-query number because it is the one that reflects what an unprompted buyer actually sees.

Retrieval state for the Charleston 2026 census: the July 2026 baseline ran without live web retrieval on ChatGPT, Claude, and Gemini. Perplexity was search-grounded at the provider level in both periods. A business not named in the July baseline was not named without live retrieval. The Sept 10, 2026 pre-shift census and the Sept 14, 2026 controlled model test both ran with live web retrieval enabled on all four models, comparing the July panel to the current default-tier model panel. See ARO™ Index Vol 3: The September 2026 Model Shift, Section 1, for the parametric-to-grounded delta and Section 3(a) for the full retrieval-condition disclosure.

What counts as selected

Matched by normalization function, not eyeballed.

A business counts as selected when a model names it unprompted in a cold market query and that name (or domain) matches the audited business, using the same normalization functions that built the entity corpus for that market. The exact functions and the full matching query are published below, so anyone can rerun it.

Matching runs on normalized name or normalized domain, never domain alone — a domain-only rule would undercount, since not every cold-query mention carries a domain.

The exact matching query · normalization functions · reruns live
Normalization functions (live in the database)
CREATE OR REPLACE FUNCTION public.phase2a_normalize_name(raw text)
 RETURNS text LANGUAGE sql IMMUTABLE AS $function$
  SELECT trim(
    regexp_replace(
      regexp_replace(
        regexp_replace(
          regexp_replace(
            regexp_replace(
              lower(trim(raw)),
              '^the\s+', ''
            ),
            '''s(\s|$)', '\1', 'g'
          ),
          '[''.,]', '', 'g'
        ),
        'barbecue', 'bbq', 'g'
      ),
      '\s+(oyster bar|real estate|restaurant|llc|inc)\s*$', ''
    )
  );
$function$;

CREATE OR REPLACE FUNCTION public.phase2a_normalize_name_aliased(raw text)
 RETURNS text LANGUAGE sql IMMUTABLE AS $function$
  SELECT CASE
    WHEN lower(trim(raw)) IN ('hall''s chophouse', 'halls chophouse') THEN 'halls chophouse'
    ELSE phase2a_normalize_name(raw)
  END;
$function$;

CREATE OR REPLACE FUNCTION public.phase2a_normalize_domain(raw text)
 RETURNS text LANGUAGE sql IMMUTABLE AS $function$
  SELECT NULLIF(
    regexp_replace(
      regexp_replace(lower(trim(raw)), '^https?://', ''),
      '^www\.', ''
    ),
    ''
  );
$function$;
Matching query (Charleston 2026 example, reruns live)

Full runnable version, including the audited-cohort definition and one documented manual alias, is published in _config/METHODOLOGY.md. Core matching logic:

select count(distinct ac.id) as selected_count
from audited_businesses ac
join selected_entities se
  on phase2a_normalize_name_aliased(ac.business_name)
     = phase2a_normalize_name_aliased(se.canonical_name)
  or phase2a_normalize_domain(ac.domain)
     = phase2a_normalize_domain(se.canonical_domain);

Where selected_entities is the deduplicated canonical-entity layer built from clean cold-query mentions (excludes off-category bleed and any retro-mined pilot data), and audited_businesses is every business with a current, public audit in the market.

Locked example: Charleston 2026

688 businesses audited · 89 ever selected in a cold buyer query · 12.9% selection rate · 1,616 clean cold mentions · 1,047 distinct entities · 43 categories · data current July 8, 2026. Full report: /research/charleston-2026.

Used in longitudinal comparisons (e.g. Model Shift reports)

Research cohorts are frozen snapshots.

A research cohort locks a list of businesses on one date, so a later comparison measures the same businesses over time. Each cohort records a membership rule, a snapshot date, and a locked count. The frozen list is authoritative. The rule describes how the list was built. It does not reproduce the list on demand.

charleston_vol1: best-known rule — audited, score version v2, live audit environment, public, 3 or more models, most recent qualifying audit before the snapshot. Snapshot Aug 26, 2026. 687 domains. Locked. The live version of this query drifts as audits accrue — 688 on July 8, 685 on Sept 13. The frozen list defines the cohort, not the live query. Running the rule today does not return 687, and it isn't supposed to.

nashville_vol1: same rule, scoped to Nashville, cutoff before July 27, 2026 — the day the Nashville July census fired. Snapshot Sept 13, 2026. 481 domains. Locked.

A cohort built after its own census already ran is disclosed as retroactively constructed with a pre-census cutoff. It is never called pre-registered.

2026-11-11 note: how cohort businesses are matched for the October to November 2026 comparison (Vol 4).

This covers charleston_vol1 (687) and nashville_vol1 (481).

  • Waves: the October wave is census_v2, run Oct 3, 2026 (UTC). The November wave is census_20261102, run Nov 3, 2026 (UTC).

Recovery fills:

  • In Nashville, some answers were filled after the main run: 1 October answer on Oct 4, November Gemini and Perplexity answers on Nov 4 to 5, and 4 November Claude answers on Nov 9.
  • Each fill used the same model and request.
  • One Nashville business is selected in November only through a Nov 9 answer.
  • What counts as selected: a cohort business counts as selected in a wave when at least one of ChatGPT, Claude, Gemini and Perplexity names it in a mention that passes the section 5b filter (source live_cold_run, off_category false). In these four batches every mention passes. The name has to match in one of three ways:
  1. Name. The entity's normalized name (phase2a_normalize_name_aliased, plus the 5c alias) equals the cohort business's normalized name. Same as section 5d.
  2. Domain. The entity's normalized domain equals the cohort domain. Same as section 5d, except on shared domains.
  3. Shared domains. 12 cohort domains are carried by more than one business in the market:
    • 3 hospital systems.
    • A hotel and the bars and spa inside it.
    • Brands with several locations where the cohort member is one location.
    • Sister shops.
    • 2 pairs of possibly different firms.

    The list was reviewed by hand and is fixed for Vol 4. On these domains a domain match counts only when the names also match. Either every word of the cohort name appears in the entity name, or every word of the entity name appears in the cohort name and the entity name is not just the brand that every business on that domain carries. A department of a hospital system counts only for that department. For example, a hospital system's urgent care is not counted for the same system's orthopaedic practice.

Entity data changes made before counting (Nov 10 and 11, 2026):

Domains:

  • 126 entity domains stored as page URLs were cut to the site's root. 34 directory or platform pages were left as they were.
  • 1,223 entities with no domain were given one where every linked mention in both waves points to the same site. Directory, media and parent-company sites were excluded, and 161 entities were skipped.
  • Duplicates: 131 sets of duplicate entities (same business, same location) were merged.

Flagged mentions:

  • The 287 flagged mentions that touch a cohort business were each linked to the entity with the same name, with a logged reason. This created 4 new hospital department entities.
  • The other 528 flagged mentions in the two waves stay unlinked and do not count.
  • Mislink: 1 mention linked to the wrong location of a multi-location brand was moved to the right location.

Effect on the counts (selected of members):

  • charleston_vol1, October: 284 of 687 (41.3%). The earlier rule gave 254 (37.0%).
  • charleston_vol1, November: 268 of 687 (39.0%). The earlier rule gave 237 (34.5%).
  • nashville_vol1, October: 218 of 481 (45.3%). The earlier rule gave 195 (40.5%).
  • nashville_vol1, November: 212 of 481 (44.1%). The earlier rule gave 181 (37.6%).
  • Why the counts rose: most of the increase is cohort businesses listed under a search-style title (for example "Nashville SEO Agency") that only a domain can match. Their entities now carry the domain.
  • Why some counts fell: 5 businesses dropped or lost a wave under the shared-domain rule. For example, a hospital-system domain no longer counts other departments.
  • Hand-check: all 79 businesses whose selection changed were checked against the names the models used.

Limitations:

Measured undercount in gained and lost businesses:

  • The opposite wave was searched for every business gained or lost between the waves.
  • 11 businesses (3 in Charleston, 8 in Nashville) were named in a wave under a name variant the rule cannot match, on an entity with no domain.
  • They are not added to the counts above. The counts are what the stated rule returns, and the 11 were found only by hand.
  • With them, the counts would be Charleston 285 and 270, and Nashville 220 and 218.

Unmeasured undercount in both waves:

  • The same kind of miss can hide a business in both waves. A business named in October and November only under name variants the rule cannot match shows as not selected in either wave.
  • That search was run only on gained and lost businesses, so this error is not measured.
  • This kind of miss can only lower the counts, never raise them.

Reproducibility:

  • The counts can be rerun only against the entity data as of Nov 11, 2026. The domain back-fill, merges, flagged-mention links and the shared-domain list are database edits and a hand-reviewed list, not published functions.
  • The edits are logged in the database, and each day's edits were preceded by a backup. Nov 10 backup: _bak_entity_fix_20261110_*. Nov 11 backup: _bak_cohort_final_20261111_*.
  • The earlier-rule counts were computed on the Nov 11 backup data, after the Nov 10 fixes, and are kept in _cohort_20261111_oldrule.

Comparability:

  • These counts replace the earlier-rule counts for this comparison only. Earlier published cohort figures are not restated.
  • The panel changed on 3 of 4 models between the two waves, so the October to November difference describes the panel change plus month-to-month churn, not market movement.
  • October reader: the name reader that pulls business names from answers ran with a shorter output limit in October (max_tokens 500, with thinking not yet disabled). Both were fixed Oct 13. If it cut long October answers short, some October names are missing. That would lower the October counts, and some businesses would look gained in November when they were not. Claude's long answers would be hit most. Truncation is neither confirmed nor ruled out. October answers were not re-read with the current reader.
V1 archive · original single-query method · through early 2026
Section 1

About the ARO Index

The ARO Index is a live market dataset tracking which businesses AI platforms actually select when buyers ask for local service recommendations by city and category. It measures selection behavior, not mentions. A business can be mentioned without being selected. ARO Index records selection.

Section 2

What We Measured

ARO Index tracked AI recommendation selection across four platforms: ChatGPT, Claude, Gemini, and Perplexity. For each market and category, a standardized query was submitted to all four platforms. The responses were recorded, analyzed, and used to produce ARO Scores and market rankings.

Rankings reflected which businesses were selected most consistently and prominently across platforms within a defined audit period.

Section 3

How Audits Worked

Audits were conducted on a rolling cycle. Each audit covered a defined city and service category. A single query was submitted in a standardized format across all four AI platforms. Results were recorded at the time of submission.

ARO Index does not rely on scraped data, third-party aggregators, or predictive modeling. All data is collected through direct platform queries conducted by the ARO Index research team.

Section 4

Audit Scope

Approved markets:

  • Charleston, SC
  • Nashville, TN
  • Atlanta, GA
  • Denver, CO
  • Miami, FL
  • Sacramento, CA
  • Fresno, CA

Additional markets are added on a rolling basis. Each market is organized by service category.

Section 5

ARO Score

The ARO Score is a composite measurement scored 0 to 100. It reflected how consistently and prominently a business was selected by AI platforms when buyers asked for recommendations in its city and category. A higher score indicates more consistent selection across platforms and queries.

Scoring methodology is proprietary. The ARO Score is a trademark of TaG Makes.

Section 6

Data Currency and Versioning

ARO Index data reflects point-in-time audit results. AI platforms update their models continuously, which means recommendation behavior can shift between audit periods. Each report references its specific audit period and dataset.

Current rankings are always available at aroindex.com.

Section 7

Research Lead

ARO Index research is conducted by Therese Grittner, founder of TaG Makes, based in Charleston, SC. Therese Grittner created the ARO Index methodology and oversees all audits, scoring, and market reporting.

Section 8

Limitations

ARO Index measures AI platform behavior at the time of each audit. Results reflect the platforms' recommendation behavior during that specific period and may not represent behavior at other times.

Sample sizes vary by market and category. Posts with samples under 100 businesses are labeled as exploratory or directional.

ARO Index does not make causal claims. All findings are observational and correlative unless otherwise stated.

Section 9

Contact

For research inquiries: therese@aroindex.com

For current rankings and market data: aroindex.com

Ask the Index×
Hi! Ask me about the ARO Score, how AI recommendation works, or what separates top-ranked businesses.