On martech and advertising buying questions, AI engines cite vendor-owned sites for 67.8 percent of the citations the Machine Relations Index can classify by source type: 485 of 715. That is the highest vendor share among the 20 categories the Index publishes on these question shapes. Editorial publications hold 12.9 percent of the same classified citations (92 of 715). The trade publication engines cite most often on martech news, martech.org, appears in 15.7 percent of the category's news runs and in 1.0 percent of its buying-question runs.
The weekly Index Report of October 19, 2026 still showed these buying shapes as collecting; only martech news had a published rate. By the October 24, 2026 release four of the category's buying-question shapes had cleared the evidence floor. This page reads them and sets them beside the other categories.
What was measured #
Every figure comes from the public release at machinerelations.ai/data/machine-relations-index.json: release mri_score_v2.1+2026-10-24+e479e9b12312 (artifact sha256 beginning e479e9b12312), generated October 24, 2026, window May 10 to October 24, 2026, 30,766 observed answer runs across six engines. The Index page for the category is Martech and Advertising.
- Buying-question shapes. Best tools (106 runs), how buyers choose (106), is it worth it (102) and problem-first research (101), 415 runs in all. Each shape has seven questions and meets the Index's evidence floor of at least 10 observed runs on at least 7 distinct run dates: the runs are 101 to 106 and the dates are exactly 7.
- Left out. Top lists and head-to-head comparisons are still collecting (76 runs on 6 dates each since October 12) and have no published rate. News-driven questions are a different kind of question and are read separately below.
- Counting. One citation here is one domain cited in one run, summed over the four shapes. A run that cites five domains adds five. The four shapes hold 1,886 such citations across 653 distinct domains.
- Classified. The Index assigns each domain one of nine source roles. The largest role, other observed source, is the unclassified layer and is not a finding about any source type: it holds 1,171 of the 1,886 citations (62.1 percent). Shares below are of the 715 classified citations unless stated. This uses the same classified-only denominator as the Index's earlier source-class study by category, which did not include martech.
- Engines. Each shape drew about 23 or 24 runs from each of ChatGPT, Claude, Gemini and Perplexity, and 4 to 6 from each of Google AI Mode and Google AI Overviews.
The result: vendors lead every buying shape #
| Source role | Citations | Share of all 1,886 | Share of 715 classified |
|---|---|---|---|
| Vendor-owned source | 485 | 25.7% | 67.8% |
| Editorial publication | 92 | 4.9% | 12.9% |
| Market and company database | 51 | 2.7% | 7.1% |
| Analyst and consulting research | 31 | 1.6% | 4.3% |
| Community and social platform | 23 | 1.2% | 3.2% |
| Wire and press-release distribution | 15 | 0.8% | 2.1% |
| Search or media platform | 14 | 0.7% | 2.0% |
| Academic and government source | 4 | 0.2% | 0.6% |
| Other observed source (unclassified) | 1,171 | 62.1% | not applicable |
Vendor-owned sources lead in each of the four shapes, and editorial publications come second in each.
| Question shape | Runs | Classified citations | Vendor-owned | Editorial |
|---|---|---|---|---|
| Best tools | 106 | 267 | 169 (63.3%) | 37 (13.9%) |
| How buyers choose | 106 | 184 | 128 (69.6%) | 28 (15.2%) |
| Is it worth it | 102 | 162 | 130 (80.2%) | 10 (6.2%) |
| Problem-first research | 101 | 102 | 58 (56.9%) | 17 (16.7%) |
The vendor share differs by shape: lowest on problem-first research (56.9 percent) and highest on is it worth it (80.2 percent), where 130 of the 162 classified citations point to vendor-owned domains as the Index classifies them.
Highest of 20 categories #
Twenty categories have all four of these shapes published. Martech and advertising ranks first on the vendor share of classified citations and first on the vendor share of all citations. The table shows the top five of the 20 and the lowest, consumer products.
| Category | Classified citations | Vendor share of classified | Vendor share of all | Editorial share of classified |
|---|---|---|---|---|
| Martech and advertising | 715 | 67.8% | 25.7% | 12.9% |
| Fintech | 1,011 | 56.0% | 12.6% | 13.9% |
| Cybersecurity | 2,259 | 53.3% | 21.5% | 20.1% |
| HR and talent | 1,347 | 50.9% | 16.2% | 20.0% |
| AI visibility and GEO | 1,969 | 48.1% | 13.0% | 15.5% |
| Consumer products | 1,960 | 0.2% | under 0.1% | 68.9% |
Fintech and cybersecurity were the two categories where the earlier study found vendor-owned sources winning every one of six shapes. That study compared six shapes and this page compares the four martech has published, so the two are not the same measurement. On the four shared shapes, martech's vendor share is 11.8 points above fintech's and 14.5 points above cybersecurity's.
One vendor does not explain it #
The largest single domain is cdp.com, which describes itself as a guide to customer data platforms and which the Index classifies as vendor-owned. It is cited in 73 of the 415 runs (17.6 percent). Removing it leaves 412 vendor-owned citations among 642 classified ones, 64.2 percent, so the lead in these four shapes does not depend on it. Fifty-nine distinct vendor-owned domains are cited in these shapes. The most cited are below, with the last two tied. The counts are observed citations of domains in answer runs. They are not rankings of products and not a recommendation.
| Domain | Runs citing it (of 415) | Share of runs | Confidence tier |
|---|---|---|---|
| cdp.com | 73 | 17.6% | B |
| cometly.com | 41 | 9.9% | C |
| improvado.io | 34 | 8.2% | B |
| bloomreach.com | 28 | 6.7% | C |
| adobe.com | 26 | 6.3% | B |
| nvecta.com | 19 | 4.6% | C |
| usermaven.com | 19 | 4.6% | collecting |
The vendor classification is recorded with its basis: 277 of the 485 vendor-owned citations belong to domains classified by review and 208 to domains classified from the site's own evidence.
The Index also records which reviewed source names appear in a question, which lets it separate citations on questions that name a vendor from citations on questions that do not, as an earlier analysis of one cybersecurity segment did for one vendor. For these four shapes it records no reviewed source named in the questions. Vendor domains outside the reviewed names are not covered by that check.
The trade press that leads news is rarely cited in the buying answers #
Martech news questions draw a different mix. The category's news-driven segment has 578 runs on 25 questions over 53 run dates. There, martech.org is the most cited domain: 91 of 578 runs, 15.7 percent. Across the four buying shapes it appears in 4 of 415 runs, 1.0 percent. Five editorial publications are cited in more buying-question runs than martech.org:
| Editorial publication | Buying-question runs (of 415) | Share |
|---|---|---|
| techtarget.com | 13 | 3.1% |
| indiatimes.com | 11 | 2.7% |
| cmswire.com | 11 | 2.7% |
| forbes.com | 9 | 2.2% |
| techradar.com | 8 | 1.9% |
| martech.org | 4 | 1.0% |
The two question sets are not the same questions, and the engine model versions differ between the segments. News questions name six vendors (6sense, Demandbase, HubSpot, Klaviyo, mParticle and Salesforce), with the questions naming each accounting for 20 to 26 runs, and the news segment draws far more runs from Google AI Mode (100) and Google AI Overviews (47) than the buying shapes do. The comparison shows that being the source engines cite on a category's news is not the same position as being a source they cite when a buyer is choosing a tool. It does not show why.
What this page does not show #
- Not why. The Index counts which domains engines cite. It does not show why an engine chose a vendor's own pages, and it does not show that any action by a vendor or a publisher moved a citation.
- Seven questions per shape. A shape describes its seven questions, not every question a marketer asks. One question accounts for up to 32.1 percent of the runs in a shape (best tools and how buyers choose), 29.4 percent in is it worth it and 28.7 percent in problem-first research. An Index analysis of October 20, 2026 measured how far rewording or rerunning a question moves cited sources.
- Seven run dates. Each published shape has 7 run dates, the minimum the evidence floor allows. Rates can move as dates accumulate. See why collecting is not zero.
- The classified layer is a minority. 62.1 percent of the citations in these four shapes sit in the unclassified layer. Vendor-owned sources are 25.7 percent of all citations, and the 67.8 percent applies only to the classified part. As classification coverage grows, either figure can change.
- No per-engine split. The six engines are pooled, and Google AI Mode and AI Overviews supply under a tenth of the runs in each shape.
- Domains, not pages. The Index does not publish which page of a domain was cited.
Reproduce it #
This command sums the same citations from the live release and prints the nine source-role totals. It reads the current release, so counts will differ after the next one; check the release id first.
curl -s https://machinerelations.ai/data/machine-relations-index.json | jq -r '[.domains[] | .source_role as $r | (.mri_score_v2.strata[]? | select(.category=="martech-advertising" and (.question_shape|IN("best_x","how_choose","is_x_worth","problem_first"))) | {r:$r, c:.runs_cited})] | group_by(.r) | map({role:.[0].r, citations:(map(.c)|add)}) | sort_by(-.citations) | .[] | "\(.role) \(.citations)"'
For the October 24 release it returns vendor_owned 485, editorial_media 92, market_database 51, analyst_research 31, community_social 23, wire_distribution 15, platform_search 14, academic_government 4 and uncategorized_source 1171.
Frequently asked questions #
Do AI engines cite vendor sites more than publications when marketers ask which martech tool to buy? In the October 24, 2026 Index release, yes: vendor-owned sites hold 485 of the 715 classified citations (67.8 percent) on martech and advertising buying questions, and editorial publications hold 92 (12.9 percent). The 62.1 percent of citations in the unclassified layer is not counted in either share.
Is martech unusual compared with other categories? On the four question shapes it has published, its vendor share of classified citations is the highest of 20 categories. Fintech is next at 56.0 percent and cybersecurity at 53.3 percent. In consumer products the vendor share is 0.2 percent and editorial publications hold 68.9 percent.
Is one vendor driving the result? No. cdp.com is the most cited single domain at 73 of 415 runs, but without it vendor-owned sources still hold 64.2 percent of classified citations, spread over 58 other vendor domains.
Does martech.org matter to AI answers? It is the most cited domain on the category's news questions, in 15.7 percent of runs, and it is cited in 1.0 percent of buying-question runs. The Index measures that pattern. It does not measure what a citation is worth to a buyer.
Where do the numbers come from?
From the public Machine Relations Index, release mri_score_v2.1+2026-10-24+e479e9b12312, published under CC BY 4.0. Quote the release id with any figure.
The short version #
- On martech buying questions, vendor-owned sites hold 67.8 percent of classified citations, the highest of 20 categories.
- Vendors lead all four published buying shapes, from 56.9 percent on problem-first research to 80.2 percent on is it worth it.
- Removing the largest vendor domain leaves the share at 64.2 percent.
- The trade publication that leads martech news is cited in 1.0 percent of buying-question runs.
- Seven questions per shape and seven run dates make this an early reading of a new category.
Machine Relations is the discipline of earning AI citations and recommendations for a brand by making that brand legible, retrievable, and credible inside AI-driven discovery. Jaxon Parrott coined it in 2024; AuthorityTech, which he founded in 2018, is the first Machine Relations agency; Machine Relations is open to anyone and machinerelations.ai is its home.