Your most loyal visitors may be reaching buyers as strangers. Buyers bid more for what they recognize, so visitors who arrive unrecognized tend to draw lower bids. If you want more revenue from the audience you still have, start with your bid request. It's the fastest place to find it, and you can check it today with tools you already have.
In my AdExchanger column on November 9, I wrote about AI agents on the buy side. They already help the largest buyers decide which inventory to evaluate. Even so, we don't expect agentic budgets to move most publishers' revenue for a few more quarters. That's no reason to wait. Getting ready for agents is a byproduct of making more money today.
The column closed with three steps for publishers. Each one protects or raises what you earn in today's auctions. Agent-readiness comes along as a side benefit. This post shows how to run each check in your own stack. We'll follow one visit through all three: identifiers, context, and access.
One visit can produce two very different bid requests
Take a visitor who comes to your site three times a week. She uses Safari and has never logged in. You know she's loyal. Her bid request may not say so.
Here is a simplified request for her visit, as it often leaves the page:
"user": { "id": "a1b2c3" },
"site": {
"domain": "example.com",
"page": "https://example.com/recipes/weeknight-pasta"
}
Here is the same visit with the gaps filled:
"user": {
"id": "a1b2c3",
"ext": { "eids": [
{ "source": "id-provider-a.com", "uids": [{ "id": "..." }] },
{ "source": "id-provider-b.com", "uids": [{ "id": "..." }] }
]},
"data": [{ "name": "example.com", "ext": { "segtax": 4 },
"segment": [{ "id": "..." }] }]
},
"site": {
"domain": "example.com",
"page": "https://example.com/recipes/weeknight-pasta",
"content": { "data": [{ "name": "example.com", "ext": { "segtax": 7 },
"segment": [{ "id": "..." }] }] }
}
Illustrative and simplified. Field names follow OpenRTB. Providers, IDs, and segments are placeholders.
The first request carries a domain, a URL, and one ID that most buyers can’t match. The second adds IDs buyers can match, content segments that classify the page, and audience segments for this visitor. When a bidder can match the visitor, it can price the impression with more confidence. That’s why recognized impressions tend to draw more bids. Not every DSP reads every ID or taxonomy, and what reaches each bidder depends on your Prebid settings and your SSPs. But no buyer can use a signal you never send, and buyer agents that evaluate inventory will be reading the same fields.
Check 1: Every visit carries an identifier that holds across your properties
Start with identity, because buyers recognize a visitor through the IDs in the request. Open one of your pages in Safari, the hardest case for IDs, and accept your consent prompt so IDs can load. Add ?pbjs_debug=true to the URL to turn on Prebid's debug logging. Then open the browser console and run pbjs.getUserIdsAsEids(). The result lists the IDs the User ID module holds for this visitor, in OpenRTB format. Some ID providers load late, so run it again after a few seconds if the list looks short. This list isn't what each bidder received. Bidder-level settings, covered in Check 3, can limit who gets each ID. To see what a specific bidder received, check its bid request in the debug output.
Repeat the test in Chrome, then on a second property you own. Look for three things:
Whether Safari visits carry any IDs at all. Many IDs depend on third-party cookies, which Safari blocks.
How far the ID count drops between Chrome and Safari.
Whether one visitor keeps one identity across your domains, or starts over on each.
If each domain mints its own ID, buyers see one loyal visitor as several strangers. We covered what that costs in “Search Traffic Keeps Falling. Your Loyal Audience Is Still There.”
Pass every ID you have consent to pass, and pass them on every visit. An ID that shows up on some visits and not others gives buyers little to recognize. Check 3 covers who should receive them.
Check 2: The request describes the visit as well as the page
Next, look at what the request says about the content and the visitor. In the same console, run pbjs.getConfig('ortb2'). This shows the first-party data you've set in your Prebid config. Segments that a data or contextual module adds at auction time may not appear here, so check the bid request in the Prebid debug output for those. Look for two blocks:
site.content.data content segments, with a segtax value naming the taxonomy that encodes them.
user.data audience segments you've defined, in the same format.
If both blocks are empty, buyers get a domain and a URL, and their bidders have to guess the rest. They price that guess into the bid.
Use the IAB's standard content and audience taxonomies where you can. Then confirm which taxonomies your SSPs and their buyers actually read. A segment a bidder can't decode adds nothing to its decision. Page context deserves its own post, and I wrote one: “Page Intelligence Is the New Premium Inventory.”
Check 3: You know which bidders can read each signal
In Prebid, bidders are the SSPs and exchanges you connect to. By default, Prebid sends every user ID to every one of them, including bidders that never win your impression but still see the request. What each SSP passes to DSPs depends on its own settings and your contract.
Open your Prebid configuration and find each ID module's bidders setting. If it's missing, that ID goes to everyone. For first-party segments, pbjs.setBidderConfig() limits which bidders receive specific data.
Then take the question to your contracts. What may a bidder keep from a request it didn't win? May it use that data to train models? The settings are quick to change. Deciding what they should be is a job for revenue and legal together. That's what owning your signal means in practice: you decide who reads it, and on what terms. Our “3-Question Addressability Audit” covers the contract side in more detail.
These checks can pay in today's auctions
None of these checks waits on an agentic standard. Each one changes how today's buyers see your inventory.
The first check works through a simple mechanism: the more IDs in a request that buyers can match, the more buyers recognize the impression, and the more bids it tends to draw. The second works the same way for content, because a bidder can price a page it can classify. The third protects the other two. A signal every bidder can keep is a signal you've given away.
In our own measurement, one publisher network ran a randomized holdout for thirteen months across more than 400 million sessions with GrowthCode enrichment. Those sessions earned 16% more Prebid revenue per session than control sessions without it. Monthly results ran from 6% to 36%, and none was negative. That's one network over one stretch of time. Your result will depend on your own traffic and stack.
GrowthCode builds the infrastructure for the first check
Publishers on GrowthCode handle the first check with two Publisher Cloud products:
Signal Graph builds a first-party identity graph you own, from signals captured on your own pages.
Signal Enrich puts that graph into the bidstream, using the 20+ universal IDs your graph maps to.
Bid enrichment is the first return on that graph. The same graph shows you which visitors are worth the most, and it can feed the CDP and personalization tools you already run.
The second check calls for contextual and segment-based enrichment, which is an upgrade to Signal Enrich. The third check stays with you. Who can read your signal is your decision, and it starts in the contract.
Want to see what the first check is worth on your own traffic? We run a 30-day POC on 3 to 5 of your sites. It takes one script and no changes to your existing stack. Request a meeting, and we'll scope it with you. If you'd rather read first, the Publisher Cloud brochure walks through how both products work.
Three questions to run this week
Does every visit carry an identifier that holds across your properties?
Does the request describe the visit as well as the page?
Do you know which bidders can read each signal?
Every "no" is a signal you aren't getting paid for, or one you're giving away.
Fix the bid request for the buyers you have today. When agentic budgets do show up, your inventory will already be easy to evaluate. If the first question comes back "no," Request a demo and we'll scope a POC on your traffic.
GrowthCode offers a 30-day POC. It’s one script, with no changes to your existing stack. You see the results on your own traffic before you commit to anything long-term.
Jonathon Shaevitz is Founder and Head of Product at GrowthCode. GrowthCode provides infrastructure to maximize signal for the digital advertising ecosystem.
