The short answer. Agentic giving is when an AI assistant researches charities for a donor, recommends one and sometimes completes the gift. A small charity does not need to buy anything to get ready. Check that AI names you correctly, make your facts identical everywhere, and make sure your donation page works from a link with the amount already filled in. The reason to start small: in the seven days to 6 October 2026, GiveReady logged 2,015 clicks on donate links, and 10 looked like people.
Want to know what AI says about your charity today? Get a free AI check.
What is agentic giving?
Agentic giving means an AI assistant acts for a donor instead of only answering a question. The donor asks where to give, the assistant compares charities and recommends one, and in some cases it carries out the donation once the donor confirms the amount and the recipient.
Fundraise Up describes it as an assistant "researching causes, comparing nonprofits, and sometimes completing the gift, often without a single website visit." Donorbox describes the same idea through Stripe: a donor's assistant searches a registry of businesses that agents can pay, finds a campaign, and finishes the gift in the chat or sends the donor to a checkout page that is already filled in.
Two parts matter to a charity, and they are separate. The first is discovery: does the assistant know you exist and describe you correctly? The second is payment: can the gift be completed from where the assistant is? Most of the work for a small charity is in the first part.
Who offers it today?
Two fundraising platforms and GiveReady offer something here, and they do different jobs. This table is based on the public pages each one publishes, read on 6 October 2026.
| How it works | Who can use it | What the page does not say | |
|---|---|---|---|
| Donorbox | Campaigns are listed in Stripe Directory. A donor's AI agent finds one, then completes the gift with Stripe Link or sends the donor to a pre-filled Donorbox page. | Organisations with a US Stripe account get early access (Donorbox blog, updated 15 September 2026). | Fees, how charities are verified, and how many gifts agents have completed. |
| Fundraise Up | Says assistants can finish a donation "based entirely on what they find published on your website". Sells an AI Visibility Assessment that tests donor questions across assistants. | Not stated. Pricing is not on the page. | Which assistants take part, and how a gift is actually completed. |
| GiveReady | A free profile that AI engines read. The Donate button sends the donor to the charity's own donation page. An on-chain USDC endpoint for agents also exists. | Any charity in the directory of 41,000+ profiles can claim its profile for free. | Nothing is sold as agent payments. That endpoint has settled 9 gifts since launch, all made by us, the last in April 2026. |
How big is it really?
Nobody has published how many gifts AI agents have completed, so the honest answer is that it is small and the evidence is thin. The figures that are public come from the companies selling the products.
Donorbox cites a study in which 4.5% of 1,728 surveyed donors use AI to find and research causes. It says traffic from AI tools to US nonprofit sites has grown 10 times since January 2024, and that the average gift from an AI-referred donor is $250. Fundraise Up says organic search traffic to nonprofit sites is down about 35% year on year, AI traffic is growing about 1,000% a year, and AI-referred donors give $250 on average against $210 from other sources.
Read those as vendor claims. Neither company shows its working, and the two Donorbox pages do not agree with each other: the explore page says AI visitors spend "4x longer" on nonprofit sites, the blog says "more than 70% longer", and the two pages name different assistants that can use the feature (OpenClaw and Grok Bot on one, Hermes, Instinct, Meta's Muse and Grok Bot on the other).
Our own logs point the same way. In the seven days to 6 October 2026, donate links on GiveReady profiles were clicked 2,015 times. Our filters classed 1,596 of those as known crawlers and 409 as scripts that walk through many charities in a row. That left 10 that looked like a person. AI assistants fetched a page while answering a real question 17 times in the last 30 days (ChatGPT 15, Claude 1, Perplexity 1). Crawlers read charity pages in the tens of thousands; people following a gift link from an AI answer are a handful.
What do AI assistants say when someone asks?
We asked Perplexity eight charity-side questions on 6 October 2026, using its API with web search on, and recorded which sites it cited.
For the six questions about getting a charity found by AI, such as "how can I get my charity found by AI" and "does my charity need llms.txt", the cited pages came from marketing agencies and nonprofit consultancies. One site, Elevation, was cited in five of the six. GiveReady was cited once, for our home page, in answer to "how do I check if AI assistants know about my charity". Our own how-to guide, published on 25 September, was cited in none of them.
The two questions about agentic giving, "what is agentic giving and should my charity get ready for it" and "can AI assistants make donations to my nonprofit", drew mostly on Fundraise Up's and Donorbox's own pages. The other sources were a few more vendors, a news site and a press release. For now, the sources answering the question are mostly the companies selling the answer.
What should a small charity do first?
Do these five things in this order, and spend money last. None of them depends on which AI assistant a donor uses.
1. Check what AI says about you. Ask ChatGPT, Google, Perplexity and Claude who your charity is and what it does, in a private window. Write down whether each names you, misses you or gets you wrong. Our guide to getting your charity found by AI has the five questions and a table to fill in.
2. Make your facts identical everywhere. Use the same name, registration number, address and one-sentence description on your website, your regulator's register, Candid or GuideStar, your fundraising platform and any directory that lists you. An assistant that finds three versions of you will either pick one at random or leave you out.
3. Say exactly what a gift pays for. Each appeal page should state what the money funds, whether it is restricted, the amounts you suggest, where you work and when the appeal ends. An assistant describing your charity to a donor can only repeat what you have written down.
4. Make your donation page work from a link. Pick the platform you already use and test it: does a link with the amount and frequency in it open a page with those already filled in? Is the receipt automatic? Does your donor system record where the gift came from? If you use Donorbox or Fundraise Up, ask them in writing what they do for agent gifts, what it costs, and how your organisation is verified. If the answers are vague, wait.
5. Check again monthly, and watch your analytics for the tag. Re-run the five questions once a month. OpenAI adds ?utm_source=openai to the links it cites, so a visit with that source in your analytics is someone arriving from a ChatGPT answer. Other assistants usually strip the referrer, so their visits look like direct traffic.
Should we pay for an agentic giving product?
Not yet, unless it costs you nothing and the provider answers four questions: which assistants can complete a gift today, how many have been completed, what it costs, and how charities are verified before they are listed. The public pages we read in October 2026 did not answer all four.
If you already use Donorbox with a US Stripe account, switching it on is a reasonable experiment, because the cost of trying looks low. If you do not, the five steps above are the same work you would do before any product helps, and they pay off whichever product wins.
How we measured this
Our logs. Click-outs are the clicks on Donate buttons on GiveReady profile pages. We separate them into known crawlers (by published user-agent), walkers (a user agent that opened donate links for five or more different charities in 30 days) and the rest. "Looked like a person" means the rest. We cannot prove each of those 10 was human. The 17 assistant fetches are the ChatGPT-User, Claude-User and Perplexity-User crawlers, which load a page at the moment someone asks a question, counted over 30 days.
Citation test. Eight questions, asked once through Perplexity's API (model sonar) on 6 October 2026. One engine, one day, so treat it as a snapshot. Claude and ChatGPT have not been asked these eight questions yet, so we make no claim about them.
The vendor pages. Read on 6 October 2026: the Donorbox explore page, the Donorbox blog post updated 15 September 2026, and the Fundraise Up agentic giving page. Their figures are theirs and we have not verified them. Product pages change, so check the current ones before relying on this table.
What we do not know. How many gifts AI agents have completed anywhere. Whether Donorbox's or Fundraise Up's products have produced any. Whether the 10 click-outs that looked like people were donors.
The live crawler counts are public at giveready.org/api/ai-traffic.
Don't want to do this by hand?
We'll ask ChatGPT, Claude and Perplexity 19 questions about your charity and email you a one-page report within two working days: whether you're named, who is named instead, and three fixes in order. Free, no account, no card.
Did AI read your charity this month?
Want us to do the check for you?
We will ask ChatGPT, Claude and Perplexity 19 questions about your charity and send you a one-page report: whether you are named, who is named instead, and three fixes in order. Ask for a free AI Visibility Audit. It is free, and it arrives within two working days. You can also claim your charity's free profile on GiveReady and correct what AI reads about you.
Written by Geordie Wardman, who runs the Finn Wardman World Explorer Fund, a small grant fund, and built GiveReady after running into this problem himself.