AI Brand Footprint Benchmark 2026 · Whole Whale
Which nonprofits does AI name when people ask who to support?
Whole Whale asked 7 AI models 255 questions a donor, volunteer or program officer would ask, across 17 causes, once from memory and once with web search, and read every answer. Explore what they found, then commission the same study for your cause.
1,785
answers from memory, and as many with web search
7
models, from GPT-4o to Llama 4
17
causes, 15 questions each
1,273
distinct organizations named from memory
By cause
Who gets named in your space
Pick a cause. Names every model gives sit in the middle of the map; a name only one model gives sits out by that model, so whether people hear it depends on which assistant they open.
On the right, the top ten from memory and with web search. Blue marks names search brings in that memory left out.
78 organizations named from memory. The top ten take 61% of mentions.
Who each model names in lung cancer
The 60 names given in at least two answers. 7 come from all seven models; 5 from just one.
Select a dot to see who names it.
From memory
What the model absorbed in training. Answers naming each, of 105.
Show as table
| Answers | Type | |
|---|---|---|
| LUNGevity Foundation | 93 | Charity, named by 7 of 7 models |
| American Lung Association | 86 | Charity, named by 7 of 7 models |
| Lung Cancer Research Foundation | 81 | Charity, named by 7 of 7 models |
| GO2 Foundation | 57 | Charity, named by 7 of 7 models |
| Lung Cancer Alliance | 53 | Charity, named by 6 of 7 models |
| Addario Lung Cancer Foundation | 49 | Charity, named by 7 of 7 models |
| Bonnie J. Addario Lung Cancer Foundation | 43 | Charity, named by 7 of 7 models |
| American Cancer Society | 43 | Charity, named by 7 of 7 models |
| Stand Up To Cancer | 32 | Charity, named by 6 of 7 models |
| International Association for the Study of Lung Cancer | 21 | Charity, named by 6 of 7 models |
With web search
What it finds and cites today. Answers naming each, of 105. 6 of these ten are not in the memory top ten (blue).
Show as table
| Answers | Type | |
|---|---|---|
| LUNGevity Foundation | 95 | Charity, named by 7 of 7 models |
| Lung Cancer Research Foundation | 70 | Charity, named by 7 of 7 models |
| American Lung Association | 60 | Charity, named by 7 of 7 models |
| GO2 for Lung Cancer | 48 | Charity, named by 7 of 7 models |
| International Association for the Study of Lung Cancer | 46 | Charity, named by 7 of 7 models |
| Lung Cancer Foundation of America | 37 | Charity, named by 7 of 7 models |
| IASLC Foundation | 34 | Charity, named by 7 of 7 models |
| Charity Navigator | 22 | Charity evaluator, named by 7 of 7 models |
| Lung Cancer Initiative | 18 | Charity, named by 7 of 7 models |
| Bill & Melinda Gates Foundation | 18 | Foundation, named by 6 of 7 models |
Competition
Some causes are open, others are locked up
In lung cancer the ten most-named organizations take 61% of all mentions. In poverty they take 26% and the field stays open.
That difference decides the job: displacing an incumbent, or getting onto a short list AI keeps repeating.
Show as table
| Cause | #1 | Top 3 | Top 10 | Organizations named | Leader |
|---|---|---|---|---|---|
| Lung Cancer | 10% | 28% | 61% | 78 | LUNGevity Foundation |
| Breast Cancer | 11% | 30% | 60% | 91 | Breast Cancer Research Foundation |
| Cancer | 8% | 18% | 47% | 131 | American Cancer Society |
| Youth Mental Health | 6% | 18% | 42% | 181 | Jed Foundation |
| Mental Health | 6% | 16% | 40% | 182 | National Alliance on Mental Illness |
| Food Insecurity | 9% | 18% | 39% | 169 | Feeding America |
| Animals | 7% | 18% | 37% | 144 | Humane Society of the United States |
| Women's Rights | 6% | 15% | 34% | 187 | Global Fund for Women |
| Arts and Culture | 5% | 13% | 33% | 164 | Mellon Foundation |
| Disaster Relief | 7% | 16% | 32% | 172 | American Red Cross |
| Human Rights | 6% | 15% | 31% | 185 | Amnesty International |
| Environment | 5% | 12% | 30% | 188 | The Nature Conservancy |
| Youth Development | 6% | 14% | 30% | 237 | Boys & Girls Clubs of America |
| Civic Education & Democracy | 4% | 10% | 28% | 180 | League of Women Voters |
| Health | 4% | 10% | 27% | 215 | Bill & Melinda Gates Foundation |
| Education | 4% | 10% | 27% | 209 | DonorsChoose |
| Poverty | 3% | 10% | 26% | 212 | Bill & Melinda Gates Foundation |
Memory vs search
The same question, asked two ways, returns two lists
The cause leader survives the switch to web search in 14 of 17 causes. Below it, only 19% of each top twenty carries over.
With search, models name fewer organizations (7.3 per answer, against 12.8 from memory), and the leader's share rises from 6.2% to 7.2%. Search rewards whoever is easiest to find and cite today.
Organizations named per answer
Show as table
| Model | From memory | With web search |
|---|---|---|
| Grok 4 | 20.2 | 11.1 |
| Kimi K3 | 17.4 | 12.2 |
| Gemini 2.5 Pro | 13.3 | 5.9 |
| Llama 4 Maverick | 11.3 | 5.0 |
| Claude Sonnet 4 | 10.5 | 5.4 |
| GPT-4o | 8.5 | 3.6 |
| Perplexity | 8.4 | 8.1 |
Sources
Charity Navigator and GuideStar are the shared referees
6 of 7 models name Charity Navigator in more than a quarter of their answers. Past that shared layer, each model has its own habits, so a current profile on the evaluators is now part of being found.
Perplexity looks source-light here only because its footnoted links weren't captured in the from-memory answers; with web search, every Perplexity answer carries sources.
Share of each model's answers naming each evaluator
| Model | Charity Navigator | GuideStar | CharityWatch | BBB Wise Giving Alliance | Candid | GiveWell |
|---|---|---|---|---|---|---|
| GPT-4o | 27% | 26% | 6.7% | 5.9% | 1.6% | 1.6% |
| Gemini 2.5 Pro | 46% | 39% | 24% | 8.6% | 26% | 19% |
| Claude Sonnet 4 | 32% | 30% | 6.7% | – | 11% | 14% |
| Grok 4 | 37% | 11% | 14% | 3.1% | 26% | 55% |
| Perplexity | 8.2% | 0.4% | 9.8% | 2.0% | 5.1% | 2.0% |
| Kimi K3 | 44% | 38% | 7.8% | 33% | 59% | 11% |
| Llama 4 Maverick | 29% | 26% | 0.4% | 16% | 2.7% | 5.5% |
Charities vs foundations
Donors and funders hear about different organizations
Of the top 100 organizations named for donor, volunteer and best-of questions, only 22 are also in the top 100 for questions about who funds the cause. Being visible to donors does not make you visible to funders.
Foundations travel further: 8 of the 10 organizations named in the most causes are foundations, while charities tend to be named in their own cause and nowhere else.
Shared by both lists
Named in the most causes
Show as table
| Organization | Kind | Causes named in | Answers naming it |
|---|---|---|---|
| Robert Wood Johnson Foundation | Foundation | 17 of 17 | 182 |
| VolunteerMatch | Charity | 17 of 17 | 73 |
| Idealist | Charity | 17 of 17 | 56 |
| Bill & Melinda Gates Foundation | Foundation | 16 of 17 | 289 |
| Ford Foundation | Foundation | 15 of 17 | 289 |
| Rockefeller Foundation | Foundation | 15 of 17 | 180 |
| MacArthur Foundation | Foundation | 14 of 17 | 146 |
| Bloomberg Philanthropies | Foundation | 14 of 17 | 97 |
| Hewlett Foundation | Foundation | 13 of 17 | 137 |
| Kresge Foundation | Foundation | 13 of 17 | 107 |
| Wellcome Trust | Foundation | 13 of 17 | 94 |
| Mellon Foundation | Foundation | 13 of 17 | 66 |
Commission an AI Footprint Study
The same method, pointed at your cause, your peers and the funders you care about: the questions your donors and program officers ask, put to the models they use, from memory and with search. You get where you stand, who AI names instead of you, the sources it trusts, and what to publish to change it.
- · A question set written for your audiences
- · Every major model, asked both ways
- · Your peers and competitors, side by side
- · A readout for your team or board
Want to track it yourself every month instead? Start free.
Source: Whole Whale, AI Brand Footprint Benchmark 2026. From-memory answers collected 2026-07-28, web-search answers 2026-08-24, both through OpenRouter. API answers are not identical to what people see in each assistant's app, and counts measure visibility in these answers, not an organization's reach, funding or effectiveness. Methods and caveats are in the full report.