AI-search gaps now flow into campaigns

AI search gaps are the buyer questions that ChatGPT, Claude, Perplexity or Google's AI features answer by citing a competitor, or by citing nobody, when they could be citing you. InsightWonder finds those gaps per brand. From this week a brand can paste its InsightWonder API key into RelayWonder, and the two products run as one loop: a gap syncs in every hour, one click turns it into a content campaign with a brief and a bounty, a partner delivers a URL, and that URL is checked until an assistant cites it. This post is the full walkthrough of the integration, including the arithmetic of what a campaign costs, how to prioritise without prompt-volume numbers, and what the integration cannot do.

Why AI search gaps belong in a partner program

Most partner content is written to the partner's own calendar. A creator reviews what they feel like reviewing, a blogger writes the comparison that was already on their list. Some of it earns commission for years; most of it is read by nobody. That is the acceptable waste of a program where the brand only pays on results.

AI assistants changed the economics of that waste. When a buyer asks an assistant "what is the best tool for X", the assistant cites a small number of pages, and those pages get the click, the signup and the renewal. If no good page exists for a question, the assistant either cites a competitor or answers from general knowledge with no citation at all. Either case is a gap, and a gap is the most precise brief a content partner can receive: here is the exact question, here is who is currently being cited, here is the format that fits. The partner program is where people who can write that page already are, and the bounty plus commission is the way to pay for it.

Setup in three minutes

1. In InsightWonder, create an API key for the brand whose questions you track. Copy it.
2. In the RelayWonder brand console, open Integrations, choose InsightWonder, paste the key. It is encrypted at rest and shown exactly once; after you save, the screen shows only the last four characters and the date added.
3. Pick which InsightWonder brand workspace maps to which RelayWonder program if you run more than one. Most brands have a single mapping.
4. The first sync runs immediately and pulls every open opportunity. After that, a sync runs every hour and pulls new and changed items only.

If the key is revoked on the InsightWonder side, the next sync fails, the integration card turns to a "key invalid" state with the time of the last successful sync, and nothing else in RelayWonder is affected. Campaigns already created keep running.

What syncs every hour

Each of the synced AI search gaps arrives with the fields we need to write a brief, and nothing we do not. The fields are:

• The buyer question, as a person would type it, plus close variants InsightWonder groups with it.
• Intent: comparison, how-to, alternative to a named competitor, pricing, integration, or category overview.
• Current state: which competitors are cited, whether the brand itself is cited, or whether no source is cited.
• Formats that fit the question. InsightWonder suggests from the same five formats RelayWonder campaigns support: youtube, tiktok, blog, newsletter, reddit.
• Last checked timestamp, so you know how fresh the state is.

One field is deliberately absent: prompt volume. InsightWonder does not estimate how many people ask a given question, and we do not invent an estimate on its behalf. We cover how to prioritise without it below.

From gap to content campaign in one click

On the AI opportunities page the brand sees the synced list, filterable by intent, state and format. Selecting one and clicking "create campaign" opens a campaign form with the brief pre-filled from the opportunity: the question as the title, the intent and cited competitors as context, the suggested formats as the allowed deliverable types, and a default instruction that the piece must answer the question directly in the first paragraph. The brand then sets three numbers.

• Bounty per delivered piece. A flat amount paid when the brand confirms delivery, regardless of later sales.
• Cap. The maximum number of pieces the campaign accepts, so that a popular question does not produce twenty near-identical posts.
• Commission. Usually the program's standard rate on sales attributed to the delivered URL through the normal tracking link and coupon code.

The campaign is published to partners in the network who have the matching formats on their profile. They apply, the brand accepts, the partner delivers a URL, the brand confirms. At confirmation the bounty is written to the ledger as a positive row and the URL is pushed to InsightWonder as a candidate citation source for that question.

The arithmetic of a campaign

Take a $120 per month product paying 20% recurring commission with a 30-day hold. The brand finds an AI search gap for the question "best invoicing tool for freelance designers", currently answered with two competitors cited. It creates a campaign with a $150 bounty, a cap of three pieces and the standard commission, allowed formats blog and youtube.

Two things stand out. The bounty is the only cost that happens before any revenue, and it is capped by design: 3 x $150 = $450, known on day one. Everything after that is commission on real invoices, so cost scales with revenue and never ahead of it. And by month six, even in this modest scenario, the three pages have attributed $5,520 of revenue against $1,482 of total cost, with commission continuing to track renewals through the append-only ledger. If none of the three pages ever gets cited and none converts, the loss is $450 and the brand still owns three pages that answer a real question.

Delivery, citation check and resolution

Once a delivered URL is pushed to InsightWonder, RelayWonder polls citation status for that question. InsightWonder re-asks the question of the assistants it tracks on its own schedule and records which sources are cited. When the delivered URL appears among the citations for the question, three things happen in RelayWonder:

1. The asset is marked cited, with the assistant name and the date of the check.
2. The opportunity moves to resolved. It stays visible with its history so the brand can see which campaign closed it.
3. The partner sees the citation on their delivered piece in the portal, which is the kind of track record other brands look at when inviting them.

If the URL is not cited after a period you choose per campaign (the default is 60 days), the asset is marked not yet cited and the opportunity stays open. The brand can accept more pieces up to the cap, change the brief, or close the campaign. Nothing about the bounty changes: it was paid for delivery, not for citation, because the partner controls the page but not the assistant.

Prioritising without prompt volume

Because there is no volume number, brands ask how to choose among forty open AI search gaps. The method we use ourselves has three steps and takes about an hour per month.

1. Sort by state. Questions where a competitor is cited are worth more than questions where nobody is cited, because the first kind is already sending buyers somewhere. Questions where you are already cited are last.
2. Sort by intent. "Alternative to [competitor]" and "[category] for [segment]" intents convert best in our experience, because the person asking is close to a decision. Category overviews convert worst. This is a pattern from our own programs, not an industry statistic.
3. Sort by fit. Check whether a partner in the network already has the format and the audience segment on their profile. A gap with a ready partner closes in weeks; a gap with no matching partner waits for recruiting.

Take the top five after sorting and run campaigns with a cap of two or three. Check the cited rate after 60 days and adjust. Our own March 2026 sample of 200 tracked questions had roughly one in five delivered pages cited within 60 days; that is a small sample from a handful of programs, so treat it as a starting expectation rather than a benchmark.

Honest limits

• No prompt volume. InsightWonder does not estimate it, so RelayWonder shows none. Any tool that shows you a volume number for AI search gaps or for an AI assistant question is estimating; ask how.
• Citation detection is as fresh as InsightWonder's last check of that question. A page can be cited on Monday and show as not yet cited until the next check.
• Assistants change their sources. A page cited in March can drop out in May. Resolved opportunities can reopen, and the history stays.
• The integration requires an InsightWonder subscription. RelayWonder's campaigns work without it; you write briefs by hand instead of from synced gaps.
• Only the citation state crosses between the products. Partner data, ledger rows and payout batches never leave RelayWonder.

What is next for the loop

The two pieces we are working on are both honest gaps today. AI drafting of the brief from the opportunity is not connected; the brief is pre-filled from structured fields and you edit it by hand. And open-web Hunter search is not live, so a gap with no matching partner in the network waits for you to invite one. Both screens say what is pending. When they land, a gap with no partner will suggest candidates, and a brief will arrive as a first draft instead of a template.