India AI Search Readiness Report: Methodology and Research Design

Research Purpose

This annual report addresses a gap in India's digital marketing intelligence landscape: while Indian businesses increasingly recognise "AI search" as a category, there is no structured, India-specific measurement of how ready Indian local businesses are to earn AI Overview citations.

The India AI Search Readiness Report provides:

• Industry-level AIO Readiness benchmarks across 8 sectors
• City-level readiness comparisons across 18 Indian metros
• Category-specific AI citation threshold data derived from observation
• Trend data comparing readiness scores year-on-year

Status: this is a research protocol, currently in progress. Nothing below reports a finding. It documents what will be measured, how, and why — see AI Search Readiness Audit if you want your own business assessed on the same 6-pillar framework today, independent of when the aggregate report publishes.

Why this report matters to a real decision

Most local-SEO budget decisions in India are still made on instinct, or on advice copied from US content that assumes Yelp instead of Practo and Google Reviews instead of JustDial. A clinic owner in Indore deciding whether to spend the next quarter's marketing budget on schema markup, a directory sweep, or another round of Google review requests is making that call blind — there's no published India benchmark telling her which of the three pillars is furthest behind for a clinic her size, in her city, in her category.

That's the decision this report exists to inform. Not "is AI search important" — every business owner already believes that by 2026 — but "which of the six pillars should I fix first, and how far behind am I compared to businesses that are already getting cited." A benchmark only becomes useful once it's specific enough to change a spending decision, which is why the report is built around industry × city cells rather than a single national number. A national average schema-adoption rate tells a Mumbai hospital chain almost nothing; a Mumbai-healthcare-specific number tells it exactly where it stands against the businesses it's actually competing with for AI Overview citations.

Why this data doesn't exist yet

Two things make an India-specific AI citation study harder to run well than it looks.

The first is measurement cost. Getting a defensible AIO Readiness Score for one business means auditing six pillars by hand — GBP fields, schema markup, review count and velocity, content structure, citation coverage across category-specific directories, and crawlability signals. None of that is scrapeable cleanly at scale; category directories like Practo, 99acres, and JustDial don't expose consistent structured data, and several fields (review velocity, response rate) require repeated observation over time rather than a single snapshot. Multiply that by 18 cities and 8 industries and the labour cost of a rigorous sample becomes the binding constraint, not the analysis.

The second is that AI Overview citation behaviour itself isn't static. Google changes which sources it surfaces, sometimes within the same week, for reasons that have nothing to do with any individual business's optimisation work. A study that runs its queries once and calls the result a "threshold" is measuring a moment, not a pattern. That's the reason this report's design commits to monthly repeat observation over twelve months rather than a single audit wave — a business that looks cited in March and uncited in April hasn't necessarily done anything wrong; the report needs enough repeated observations to tell the difference between noise and a real signal.

The 6-Pillar AIO Readiness Score

Pillar 1 — Entity (0–20 points):

Sub-dimensions:

• GBP verification status (0–5): verified = 5; unverified = 0
• GBP category precision (0–5): most specific applicable = 5; broad category = 1–3
• GBP completeness (0–5): all fields complete = 5; sparse = 1–2
• NAP consistency across top 5 directories (0–5): fully consistent = 5; 2+ inconsistencies = 0–2

Pillar 2 — Schema (0–20 points):

Sub-dimensions:

• LocalBusiness schema presence (0–8): correct subtype + all required fields = 8; absent = 0
• FAQPage schema presence (0–6): present and valid on service pages = 6; absent = 0
• sameAs links completeness (0–3): 5+ directory profiles linked = 3
• HowTo or Speakable schema (0–3): present on relevant pages = 3

Pillar 3 — Reviews (0–20 points):

Sub-dimensions:

• Review count vs category threshold (0–8): above threshold = 8; at threshold = 4; below = 0–3
• Average rating (0–4): 4.5+ = 4; 4.0–4.4 = 3; 3.5–3.9 = 2; below 3.5 = 0
• Monthly velocity (0–4): above category benchmark = 4; at benchmark = 2; below = 0–1
• Response rate (0–4): 90%+ = 4; 70–89% = 2; below 70% = 0–1

Pillar 4 — Content (0–20 points):

Sub-dimensions:

• Homepage answer capsule (0–4): direct entity statement in first 60 words = 4
• Service page answer capsules (0–4): direct answer in first 60 words on 3+ service pages = 4
• FAQ sections on service pages (0–6): FAQ with 5+ questions on 3+ service pages = 6
• E-E-A-T signals (0–6): author credentials, accreditations, team page = 6

Pillar 5 — Citations (0–20 points):

Sub-dimensions:

• Category-primary directory presence (0–8): Practo/Zomato/99acres complete = 8
• JustDial presence (0–4): complete listing = 4
• Social media business pages (0–4): Facebook + LinkedIn complete = 4
• Industry association listings (0–4): IMA/NABH/CREDAI/ICAI as applicable = 4

Pillar 6 — Crawlability (0–20 points):

Sub-dimensions:

• AI bot access in robots.txt (0–8): GPTBot + PerplexityBot explicitly allowed = 8
• llms.txt presence (0–4): present and complete = 4
• Page speed (Core Web Vitals) (0–4): pass all = 4
• Mobile usability (0–4): no GSC mobile errors = 4

Total: /120 → Normalised to /100

Research Design

Sample: Angryturtle manages 143+ GBPs across 8 industries. The research extends beyond managed clients through a sample audit programme — businesses across each industry and city that are not Angryturtle clients are audited using the same 6-pillar framework to generate industry benchmarks.

Data collection protocol: Each business in the research sample undergoes a structured 6-pillar audit conducted by trained Angryturtle analysts using the standardised AIO Readiness Audit framework described above.

Audit cadence: Annual full audit of the complete sample. Quarterly partial audits of a rotating subsample to capture mid-year changes.

What a rigorous version of this study requires

A benchmark that businesses will actually use to make decisions has to survive three checks, and each one shapes the design above.

It needs a sample large enough per cell that a single unusual business doesn't distort the average. An industry × city cell with three businesses in it produces a number that swings wildly if one of the three happens to be an outlier; that's why the design scales the sample per cell rather than reporting national averages dressed up as local ones.

It needs auditor consistency. A 6-pillar rubric is only as good as the agreement between two people scoring the same business — if two trained analysts scoring the identical GBP profile land 20 points apart on a 120-point scale, the score isn't measuring the business, it's measuring the auditor. That's why inter-rater reliability testing on a rotating 10% subsample runs continuously rather than once at launch.

And it needs a defined observation window before any threshold or benchmark gets published, because a single snapshot can't distinguish a stable pattern from a fluke. See the Review Velocity & AI Citation Threshold Study for the twelve-month panel design this same logic drives on the reviews side specifically.

What would invalidate a benchmark like this

A few things could make the eventual numbers misleading rather than useful, and they're worth naming honestly rather than glossing over.

Selection bias is the biggest one. If the non-client sample audit programme ends up disproportionately auditing businesses that are easy to find and easy to audit — larger, more digitally active businesses — the "industry average" will overstate how ready the typical small business actually is. Stratified sampling by business size, not just by industry and city, has to be part of the correction, and the published methodology will disclose the size distribution actually achieved, not just the target.

Platform drift is the second. Google, ChatGPT, and Perplexity all update how they select and display AI answers on their own schedules, unconnected to anything an Indian business does. A benchmark measured in March and one measured in September could differ because the underlying platforms changed, not because Indian businesses got more or less ready. The annual cadence with quarterly partial refreshes exists specifically to catch and flag this kind of platform-driven shift rather than let it masquerade as a readiness trend.

And commercial conflict of interest has to be managed structurally, not just disclosed. Angryturtle sells AEO services; a study run by a commercial provider needs the non-client audit sample to be genuinely blind to commercial relationship, which is why that separation is written into the data integrity protocol below rather than left as an assumption.

How a business can run a scaled-down version of this itself

A single business doesn't need 1,000 samples to learn something real — it needs its own score and three or four comparable competitors' scores, run consistently.

Start with the six pillars above and score your own GBP profile honestly against the sub-dimension descriptions — category precision, completeness, review count against what you can observe from competitors ranking in your city for your category, schema presence (a free structured-data testing tool will show you this in under a minute), and crawlability (check your robots.txt for GPTBot and PerplexityBot access directly). Then pick the three businesses currently ranking above you in the local pack for your primary category + city query and score them the same way, using only what's publicly visible.

The output isn't a validated AIO Readiness Score — it's a directional gap map: which of the six pillars is furthest behind the businesses actually winning the visibility you want. That's often enough to decide where next month's effort goes, without waiting for a published benchmark. Angryturtle's AI Search Readiness Audit runs the same framework professionally, with the review-velocity and directory-coverage checks that are hard to do manually, if you want the comparison done properly against real citation observation rather than self-assessment.

AI Citation Threshold Research Methodology

One of the report's most distinctive contributions is empirically derived AI citation thresholds — the review counts at which businesses in each category and city are observed to begin receiving AI Overview citations.

Methodology: For each category × city combination in the sample:

1. Run target queries ("best [category] in [city]") in Google AI Overviews monthly
2. Record which businesses appear in AI Overviews
3. Query those businesses' Google review counts
4. Identify the minimum review count among AI-cited businesses in each category × city
5. Track this minimum threshold monthly across 12 months

The observed threshold is the minimum review count that has been observed earning AI Overview citations in that category × city during the research period. This is an observed floor, not a guarantee — businesses above the threshold are not guaranteed AI citations; businesses below are very rarely observed earning them. There is no published, universal review-count number that unlocks AI citation; review signal behaves as a gradient relative to your local competitive set, not a fixed gate.

What US-market research does and doesn't tell us

Local-SEO literature from the US market has documented review-count and schema-adoption patterns for years, but none of it transfers cleanly to Indian AI search behaviour. Practo functions for Indian healthcare the way no single US directory does for any category; JustDial's legacy data coverage has no real US equivalent; Hindi and regional-language query volume changes what "vernacular AEO" even means in a way that doesn't map onto any bilingual US market. Citing a specific US study's numbers as if they applied to an Indian dermatology clinic in Koramangala would be a category error, not just an approximation — which is exactly why this report is being built from Indian observation rather than adapted from an existing Western dataset.

Data Integrity Protocols

Observer consistency: All audits conducted using the same standardised rubric. Inter-rater reliability testing on 10% of audits (two auditors independently scoring the same business) with variance < 5 points on the 120-point scale.

No fabricated statistics: All research findings are based on observed data from the audit sample and from direct AI Overview query observation. No statistics are extrapolated beyond the sample without appropriate confidence interval disclosure.

Conflict of interest disclosure: Angryturtle is a commercial provider of AEO services. Research conducted on non-client businesses is conducted without knowledge of their commercial relationship with Angryturtle or competitors. Managed client data is included in aggregate only, without identifying individual client results.

Applications of the Research

Industry benchmarking: "What AIO Readiness Score does a healthcare clinic in Bengaluru need to be above average in its category?" — answerable from the report's industry × city benchmark data. See how this connects to broader entity SEO and knowledge graph work for the Entity pillar specifically.

Investment prioritisation: "Which pillar has the lowest average scores across Indian healthcare businesses?" — directs investment to the highest-opportunity gap. The schema markup pillar and the GBP optimization pillar are the two most common starting points once a business sees its own gap map.

Threshold calibration: "How many reviews does my dental clinic in Jaipur need to start appearing in AI Overviews?" — answerable from the threshold research data, alongside the broader context in reviews and AI search visibility.

How and when findings will be published

Aggregate industry-level and city-level findings will be published as part of the annual India AI Search Readiness Report once the twelve-month observation window and inter-rater validation are complete. Quarterly partial-sample refreshes will surface directional updates between annual publications, clearly labelled as partial-sample and not the full validated benchmark. Individual business scores stay confidential; only aggregated cells with adequate sample size get published, consistent with the data integrity protocol above. Businesses tracking AI search visibility more generally, or specifically watching the AI Overviews landscape for local business in India, are the natural first audience for the report once it's live.

FAQ Section

Q: How is this research different from generic AI search reports? A: Existing AI search reports (primarily from US-based firms) don't address India-specific signals: Practo, JustDial, Zomato, vernacular search, NMC compliance, or Indian city-level threshold variation. This is the only structured AI search readiness research specific to the Indian local business context.

Q: How can my business participate in the research? A: Angryturtle conducts the audit sample selection annually. Businesses interested in participating in the research can indicate interest through the Angryturtle website. Participating businesses receive their individual AIO Readiness Score as part of participation.

Q: Is the research data made public? A: Industry-level and city-level aggregate findings are published publicly. Individual business data is confidential. Managed client data appears in aggregate only.

Q: Can I get my own AIO Readiness Score before the report publishes? A: Yes. The AI Search Readiness Audit runs the identical 6-pillar framework on a single business at any time, independent of the annual report's publication schedule.

Get your business's AIO Readiness Score audited →

Internal links: AI Search Readiness Audit · AEO Services · Share of AI Voice Tracking · Share of AI Voice India Study · Review Velocity & AI Citation Threshold Study · Entity SEO & Knowledge Graph blog · Schema Markup for AI Search blog · GBP Optimization for AI Search blog · Reviews & AI Search Visibility blog · Measuring AI Search Visibility blog · AI Overviews & Local Business in India blog · GBP Management Services · Product: Rank OS · Product: Ask Maps / AIO · Research hub