---
title: "Which of 238 AI and automation agencies do AI engines cite? 24 of them take 78% of the citations"
description: "238 AI and automation agencies measured: 67.6% get cited by AI engines, but the top 24 take 78% of citations. Data, method and CSV."
url: "https://theinternetninja.com/recon/agency-ai-citations-2026/"
updated: "2026-09-27"
---

# Which of 238 AI and automation agencies do AI engines cite? 24 of them take 78% of the citations

> We screened 615 directory listings from 7 partner directories, kept the 238 that sell AI or automation work as a primary offering, and counted how often AI platforms cite each one (Ahrefs AI index, 2026-09-26). 161 of 238 are cited at least once, but the median agency is cited 3 times, and the top 24 hold 78.4% of all 29,717 citations. Citations line up with search authority far more than with anything on the agency's own case-study pages.

- Study type: Recon measurement by The Internet Ninja. TIN is the instrument; this is a measurement, not an adjudication of any claim.
- Edition: 2026-09-27
- n = 238 of up to 500
- Population: Standalone websites of agencies that state AI or automation implementation or consulting as a primary offering (headline or first listed service), live on 2026-09-27. Screened in each directory's listed order from Workato, Zapier, n8n, Airtable, Tray.ai, Clutch and UiPath partner listings; duplicates merged by root domain. (from 615 screened listings; full screening record in recon/agency-ai-citations-2026/population.csv)
- Pre-registered: method file `recon/agency-ai-citations-2026/method.md` committed at [`71b53deedf80`](https://github.com/ManeeshBehera/plan-2k27-1M/commit/71b53deedf8015201e63217f76cf2aefb24d4f21) before collection
- Dataset: [data.csv](https://theinternetninja.com/recon/agency-ai-citations-2026/data.csv), 238 rows, one per entity (CC BY 4.0)
- Population file: [population.csv](https://theinternetninja.com/recon/agency-ai-citations-2026/population.csv), 615 screened sites
- Canonical URL: https://theinternetninja.com/recon/agency-ai-citations-2026/

## Key findings

| # | Finding | Value | n |
| --- | --- | --- | --- |
| 1 | 161 of 238 agencies were cited at least once by an AI platform, but the median agency was cited 3 times | 67.6% | 238 |
| 2 | The 24 most-cited agencies hold 78.4% of all 29,717 AI citations | 78.4% | 238 |
| 3 | 173 of 238 agencies were never cited by ChatGPT | 72.7% | 238 |
| 4 | Cited agencies have a median Domain Rating of 41; agencies with no citations, 9 | 41 vs 9 | 238 |
| 5 | Ranking keywords and AI citations rise together (Spearman 0.92) | 0.92 | 238 |
| 6 | Only 22 of 168 first-listed case studies name the client, give a number and show a date | 13.1% | 168 |
| 7 | Among agencies of similar Domain Rating, a case-study page shows no consistent gap in citation rate (exploratory) | gap closes | 235 |

Every agency on this page sells AI or automation work, so every one of them has a reason to want AI engines to mention it. We wanted to know how many actually get cited, how unevenly those citations fall, and what the cited agencies have in common that the others do not. So we built the population first, wrote the method down before collecting anything, and measured all 238 agencies on the same day.

## 1. Most agencies are cited, but barely

We counted citations across all AI platforms in the Ahrefs AI index (AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, Copilot, Grok), as of 2026-09-26.

| Citations (all platforms) | Agencies | Share |
|---|---|---|
| 0 | 77 | 32.4% |
| 1 to 9 | 58 | 24.4% |
| 10 to 99 | 54 | 22.7% |
| 100 to 999 | 45 | 18.9% |
| 1,000 or more | 4 | 1.7% |

The mean is 124.9 and the median 3: a handful of very large counts pull the average far above what a typical agency gets, which is why this page reports medians.

## 2. A small group takes almost everything

The 24 most-cited agencies (the top 10%) hold 78.4% of all citations. The top five alone hold 43.8%.

| Rank | Agency | AI citations | ChatGPT | AI Overviews | Domain Rating | Size (as the directory states it) |
|---|---|---|---|---|---|---|
| 1 | accenture.com | 5,598 | 414 | 170 | 91 | not stated |
| 2 | cognizant.com | 2,536 | 178 | 63 | 82 | not stated |
| 3 | hcltech.com | 2,169 | 164 | 60 | 81 | not stated |
| 4 | incentro.com | 1,779 | 5 | 189 | 56 | not stated |
| 5 | eidebailly.com | 932 | 9 | 9 | 70 | not stated |
| 6 | genpact.com | 866 | 60 | 30 | 77 | not stated |
| 7 | dxc.com | 836 | 58 | 35 | 80 | not stated |
| 8 | valantic.com | 749 | 31 | 42 | 75 | not stated |
| 9 | reply.com | 741 | 30 | 48 | 79 | not stated |
| 10 | softwaremind.com | 729 | 3 | 32 | 59 | 1,000+ |

Most of the top of the list is large IT-services firms that list AI first on their homepage. That is a feature of the population rule, not an editorial choice: they qualify on the same mechanical test as a five-person n8n shop.

## 3. ChatGPT is the hardest room to get into

| Platform | Agencies never cited | Share of that platform's citations held by the top 24 |
|---|---|---|
| All platforms | 77 of 238 | 78.4% |
| ChatGPT | 173 of 238 | 92.7% |
| Google AI Overviews | 140 of 238 | 76.9% |

## 4. AI citations follow search authority

We fixed four off-site measures in advance and compared them with AI citations. All four move with citations; ranking keywords most closely.

| Measure | Median, cited agencies | Median, uncited agencies | Spearman with AI citations |
|---|---|---|---|
| Ranking keywords | 41 | 0 | 0.92 |
| Organic traffic (Ahrefs estimate) | 286 | 0 | 0.86 |
| Domain Rating | 41 | 9 | 0.81 |
| Referring domains | 1,367 | 729 | 0.78 |

n is 161 cited and 77 uncited. These are rank correlations across agencies on one date: they say which agencies tend to have both, not that one produces the other.

## 5. Agency case studies are common, and thin

For every agency we opened the first case study in its own listed order and checked three things a buyer (or a model) needs to trust it: a named client, a number with a unit, and a date on the page. 235 of 238 sites could be read; the rest were unreachable or hid their list behind scripts.

| Item on the first-listed case study | Present | Share |
|---|---|---|
| Has a case-study page at all | 170 of 235 | 72.3% |
| Names the client | 102 of 168 | 60.7% |
| Gives a number with a unit | 128 of 168 | 76.2% |
| Shows a date on the page | 40 of 168 | 23.8% |
| All three | 22 of 168 | 13.1% |

The date is the item most often missing. A result with no date cannot be placed in time, which matters for anything a reader might try to reuse.

## 6. Case studies and citations, compared like for like (exploratory)

Agencies with a case-study page are cited more overall: 128 of 170 (75.3%) against 30 of 65 (46.2%), with median citations of 15 and 0. But bigger, stronger domains are also more likely to publish case studies. Splitting the 235 readable agencies into three Domain Rating bands of roughly equal size:

| Domain Rating band | Cited, with a case-study page | Cited, without one |
|---|---|---|
| Low (DR up to 18) | 12 of 41 (29.3%) | 13 of 41 (31.7%) |
| Middle (DR 18 to 41) | 44 of 56 (78.6%) | 14 of 21 (66.7%) |
| High (DR above 41) | 72 of 73 (98.6%) | 3 of 3 (100.0%) |

Within each band the gap is small, and in the lowest band it runs the other way. This check was not in the pre-registered plan, so it is labelled exploratory: it is a reason not to read finding 5 as advice, not a result on its own.

## 7. Size and directory

| Size band (as stated by the directory) | Agencies | Cited at least once | Median citations |
|---|---|---|---|
| 1-9 | 49 | 22 of 49 (44.9%) | 0 |
| 10-49 | 22 | 14 of 22 (63.6%) | 12 |
| 50-249 | 16 | 16 of 16 (100.0%) | 100 |
| 250-999 | 6 | 6 of 6 (100.0%) | 215.5 |
| 1,000+ | 3 | 3 of 3 (100.0%) | 348 |
| not stated | 142 | 100 of 142 (70.4%) | 4 |

| Listed by | Agencies | Cited at least once | Median citations | Median Domain Rating |
|---|---|---|---|---|
| Workato | 91 | 65 of 91 (71.4%) | 4 | 25 |
| Clutch (AI list) | 30 | 28 of 30 (93.3%) | 100 | 51 |
| n8n | 29 | 18 of 29 (62.1%) | 2 | 26 |
| Zapier (Gold, Silver) | 29 | 12 of 29 (41.4%) | 0 | 6 |
| Airtable | 20 | 14 of 20 (70.0%) | 3 | 28.5 |
| Zapier (Platinum) | 16 | 6 of 16 (37.5%) | 0 | 14.5 |
| Tray.ai | 13 | 9 of 13 (69.2%) | 3 | 33 |
| UiPath | 10 | 9 of 10 (90.0%) | 103 | 48 |

Groups with fewer than 10 agencies are shown but not interpreted. The directory an agency came from mostly tracks its size: Clutch and UiPath listings skew to larger firms; Zapier and n8n listings skew to small specialist shops.

Clutch's list order can reflect paid placement, so we recorded Clutch's own Premier label for every Clutch listing. 26 of the 30 Clutch agencies carried it; with only 4 without it, the comparison is too small to read.

## 8. Agencies cited above their weight

Across the 155 agencies with a Domain Rating under 40, the median agency has 1 AI citation. These eight sit far above that. They are worth studying one by one, which this dataset alone cannot do.

| Agency | Domain Rating | AI citations | Organic traffic (Ahrefs estimate) |
|---|---|---|---|
| axisconsulting.io | 31 | 313 | 280 |
| consultevo.com | 36 | 245 | 308 |
| sketchdev.io | 37 | 167 | 1,743 |
| 3pillar.ai | 39 | 136 | 2,034 |
| wemakefuture.com | 37 | 73 | 592 |
| ikconsulting.com | 12 | 72 | 241 |
| prowesssoft.com | 24 | 66 | 969 |
| ajackus.com | 31 | 48 | 299 |

## What this suggests if you run an agency

- **Treat AI citation as a search-authority problem first.** Nothing else we measured tracked citations as closely as ranking keywords and Domain Rating.
- **Expect ChatGPT to be the last place you show up.** 96 of the 161 agencies cited somewhere had no ChatGPT citations at all.
- **Date your case studies.** Three in four first-listed case studies carry no date. It is the cheapest of the three items to fix, and the one a reader needs to trust a result.
- **Do not expect case-study pages alone to move the needle.** Compared like for like, they did not separate cited from uncited agencies.

These are readings of an association across agencies on one date. None of them is a tested intervention.

## How this was measured

1. **Method first.** The method was committed on its own before any data was collected, then amended three times before collection (population sources, a mechanical headline rule, coverage corrections), each amendment with its reason.
2. **Population.** 615 listings from seven directories were screened in each directory's own order against a written rule: a standalone agency site, AI or automation in the homepage headline or first listed service, live on the day. 238 qualified. The full screening record, including every exclusion and its reason, is in the repository.
3. **Screening check.** An independent second screener, blind to the first calls, re-screened a random 62 rows and agreed on 60 (96.8%, Cohen's kappa 0.93). A separate earlier screen by TIN's research agent agreed with ours on 32 of 35 overlapping sites.
4. **Freeze.** The population was committed and frozen before the first Ahrefs call.
5. **Collection.** Ahrefs AI citation counts (all platforms, ChatGPT, AI Overviews) and batch metrics for all 238 agencies, index date 2026-09-26, collected 2026-09-27. 17,884 Ahrefs units, logged per call and reconciled against the account meter.
6. **Case-study checklist.** Coded from each site's first-listed case study under a fixed protocol; a second coder, blind to the first, re-coded a random 24 sites; on the 23 both could read, they agreed on 71 of 74 item-level judgments (95.9%).
7. **Snapshots.** Every agency's homepage was captured on the day of collection (text plus a hash of the raw HTML), with the nearest Wayback Machine capture where one exists.

## Method record

Amendments, each landed before any outcome data:

- `recon/agency-ai-citations-2026/amendment-v1.1.md` at [`8ee393d7a81f`](https://github.com/ManeeshBehera/plan-2k27-1M/commit/8ee393d7a81f359b5fde969cf74f376f9be21e32)
- `recon/agency-ai-citations-2026/amendment-v1.2.md` at [`9370108143f0`](https://github.com/ManeeshBehera/plan-2k27-1M/commit/9370108143f06632c65aa4e6c39e74a9b6f87206)
- `recon/agency-ai-citations-2026/amendment-v1.3.md` at [`5811c6e55582`](https://github.com/ManeeshBehera/plan-2k27-1M/commit/5811c6e55582feb19f9caf9b95075ce68462c8ed)

| Metric | Unit | Tool | Computation |
| --- | --- | --- | --- |
| AI citations, all platforms | count | Ahrefs site-explorer-ai-responses-count (all_platforms), mode subdomains, country unset | Citation links to the agency's domain in AI answers (AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, Copilot, Grok), Ahrefs AI index as of 2026-09-26. |
| ChatGPT citations | count | Ahrefs site-explorer-ai-responses-count (chatgpt) | Citation links in ChatGPT answers, same index date. |
| Google AI Overviews citations | count | Ahrefs site-explorer-ai-responses-count (google_ai_overviews) | Citation links in Google AI Overviews, same index date. |
| Domain Rating | 0-100 | Ahrefs batch-analysis | Ahrefs Domain Rating of the root domain, subdomains mode. |
| Referring domains | count | Ahrefs batch-analysis | Unique domains linking to the agency domain. |
| Organic traffic (Ahrefs estimate) | visits per month | Ahrefs batch-analysis | Ahrefs estimate of monthly organic search visits. An estimate, not a measurement of real traffic. |
| Ranking keywords | count | Ahrefs batch-analysis | Keywords the domain ranks for in the top 100 organic results. |
| Has a case-study page | 0 or 1 | Page read, recon/agency-ai-citations-2026/checklist-protocol.md | 1 if the site lists at least one client engagement write-up. |
| Case-study completeness | 0-3 | Page read, same protocol | On the first-listed case study: named client + a number with a unit + a visible date, one point each. |

Design: correlational. The findings report what the measured entities have in common, never that a trait produced an outcome.

## What this study cannot say

- Citation counts come from one index (Ahrefs) on one date. They are a sample of AI answers Ahrefs tracks, not every answer every engine gives, and they move week to week.
- The population is agencies listed in seven partner directories, screened in each directory's own order. Make, GoodFirms and Zapier listings behind bot challenges could not be read, and 32 sites that served a bot challenge were excluded and counted. It is not every AI or automation agency.
- Large consultancies (Accenture, Cognizant, HCLTech and others) qualify when AI or automation is their headline or first listed service, so the population mixes firms of very different size. Size bands are what the directories state, and most directories state none.
- The design is correlational and cross-sectional. Nothing here shows that raising Domain Rating, rankings or case-study quality raises AI citations.
- Organic traffic is an Ahrefs estimate, not measured traffic.
- The case-study checklist reads only the first-listed case study on each site, in listed order. A site's best case study may be elsewhere.
- Finding F7 is exploratory: it was not in the pre-registered analysis plan and is reported as a check on finding F6's context, not as a result.
- The method was amended three times (v1.1 to v1.3) and the population frozen before any Ahrefs call. Each amendment and its reason is in the repository.

## Questions this study answers

**How many AI and automation agencies do AI engines actually cite?**

In this study, 161 of 238 (67.6%) were cited at least once across AI platforms in the Ahrefs AI index as of 2026-09-26. The median agency was cited 3 times, and 77 were not cited at all.

**Does ChatGPT cite automation agencies?**

Rarely. 173 of 238 agencies had no ChatGPT citations at all, and the top 24 agencies held 92.7% of all ChatGPT citations in the dataset.

**What do the most-cited agencies have in common?**

Search authority. AI citations rank-correlate with ranking keywords (Spearman 0.92), organic traffic, an Ahrefs estimate (0.86) and Domain Rating (0.81). This is an association across 238 agencies on one date, not evidence that raising any of them raises citations.

**Do case studies help an agency get cited by AI?**

Agencies with a case-study page were cited more often overall, but among agencies of similar Domain Rating the gap was small and inconsistent. Only 22 of 168 first-listed case studies named the client, gave a number and showed a date.

**Can I download the data?**

Yes. Every agency, every metric and its observation date is in the CSV linked at the top of this page, along with a snapshot reference for what was measured.

## Related

- [Generative engine optimization: what the controlled evidence actually shows in 2026](https://theinternetninja.com/blog/generative-engine-optimization-what-the-controlled-evidence-shows-2026/)

## Editions and errata

This edition: 2026-09-27. A refresh republishes on the same URL, and every prior edition's dataset stays downloadable.

No erratum.