AI & the open web

Time Is Now Selling Ads to AI Bots. Your Crawl Just Became Inventory.

APH Publisher Desk ·5 min read Share Print
In this piece
    Figure The four points the piece sets out Why this matters for publishers →
    1. 01

      You are already operating a second audience — unpriced.

    2. 02

      Crawl just got a public reference price.

    3. 03

      The measurement vacuum is a risk and an opening.

    4. 04

      Structure is becoming a commercial asset.

    Publisher Desk

    Every publisher argument about AI crawlers to date has been framed as defence: block them, license them, sue them, or eat the loss. This week Time opened a fourth option nobody had seriously operationalised — sell to them.

    Time has converted its pages to markdown and, working with the AI advertising firm Mobian, begun serving sponsored FAQ-style blocks aimed at AI agents rather than human readers. The units are targeted contextually or by date range, and measured on three currencies no human campaign uses: visibility, favourability and accuracy — in effect, whether the AI systems consuming the page see the sponsor’s message, represent it warmly, and get it right. Ally Bank and the Project Management Institute are among the first buyers. And the underlying arithmetic is the striking part: Time now sees more bot traffic than human traffic on most days, and says the new inventory prices at a premium to comparable human-facing placements.

    COO Mark Howard’s framing does the most work in the fewest words: “This is a growing traffic source, and therefore a growing source of inventory.” That single sentence reclassifies the crawl. For three years publishers have treated agent traffic as a cost line at best and theft at worst. Time is the first major publisher to treat it as an audience — one that arrives daily, reads everything, and increasingly stands between publishers and the humans who used to click through.

    01What is genuinely new here — and what isn’t proven

    The reframe is genuinely new. The market is not — yet. Two named buyers is a pilot, not a category, and it is worth being precise about which parts of this announcement are load-bearing.

    What’s real: the traffic. Bot-majority days are not a Time anomaly; any publisher who splits server logs by user agent will find agent request volumes that have grown far faster than human sessions. That audience exists whether or not anyone prices it. Also real: the format logic. Agents parse structure, not display ads — a machine-readable FAQ block is a plausible unit for a consumer that reads markdown and ignores creative.

    What’s unproven: everything downstream of the sale. The measurement currency — visibility, favourability, accuracy — has no accepted standard, no accreditation, and no independent verification path today. A brand buying “favourability” in AI answers is buying a metric its vendor defines and its vendor measures. The premium pricing is Time’s characterisation, not an audited yield figure. And nobody yet knows whether influencing an agent’s intermediate reading of a page reliably changes what the agent tells the human at the other end — which is the outcome the advertiser is actually paying for.

    None of that makes the move wrong. It makes it early. The publishers who did well out of early header bidding were not the ones who copied the first implementation; they were the ones who used the signal to get their own infrastructure ready while the standards caught up.

    02The infrastructure pays off regardless

    Here is the operator’s read: you do not need to believe in bot advertising to act on this story, because every prerequisite for selling to agents is also a prerequisite for licensing to them or blocking them.

    To sell to agents, you need to know which agents visit, how often, and against which content — that’s log analysis by user agent and ASN. To license, you need the same data, because a licensing negotiation without crawl telemetry is a negotiation without a meter. To block, you need per-crawler controls and logged access, or your block is a gesture. The same stack — measurement, gating, structured content — serves all three strategies. The only losing position is the current default at most publishers: unmeasured, ungated crawl given away free while the humans it used to bring stop arriving.

    Time's pilot may or may not become a market — two buyers and a vendor-defined metric is a thin foundation, and the honest read is that nothing here is proven beyond the traffic itself.

    03Why this matters for publishers

    You are already operating a second audience — unpricedIf Time sees bot-majority traffic most days, mid-size publishers are living some version of the same curve. That audience consumes your server capacity and your content today, and returns nothing unless you build the counter, the gate and eventually the rate card.
    Crawl just got a public reference priceEven as a pilot, Time's premium claim changes negotiations. The next AI company that wants free access to your content can now be told that a major publisher charges for agent attention — the "it's just crawling" framing has lost its innocence.
    The measurement vacuum is a risk and an openingVisibility, favourability and accuracy are vendor-defined metrics with no standard behind them. Early sellers will help write that standard; late sellers will inherit whatever the early ones and their vendors agreed.
    Structure is becoming a commercial assetTime converted its pages to markdown because agents parse structure. Clean, machine-legible page architecture is quietly turning into sales infrastructure — the same way ads.txt and sellers.json became table stakes for human-facing demand.

    04What publishers should do

    05The bottom line

    Time’s pilot may or may not become a market — two buyers and a vendor-defined metric is a thin foundation, and the honest read is that nothing here is proven beyond the traffic itself. But the reframe is permanent. Crawl is a traffic source; traffic sources get rate cards; and the publisher default of giving agents unmetered, unpriced access now looks like what it always was — a subsidy to the best-funded companies on earth. Whether you end up selling to bots, licensing to them, or shutting them out, the work is the same and it starts in your server logs. Time just made “we don’t know our bot share” an unacceptable answer.

    Sources & caveats

    Sources: Digiday, “Time has started serving ads to AI bots” (Sara Guaglione, 30 July 2026), for the Mobian partnership, the markdown conversion, the sponsored FAQ format, the targeting and measurement approach, the Ally Bank and Project Management Institute buys, Mark Howard’s quote and the bot-versus-human traffic claim. Note the caveats: the bot-traffic share and premium pricing are Time’s own characterisations, not audited figures; the visibility/favourability/accuracy measurement currency is vendor-defined with no industry standard behind it; and with two named buyers this remains a pilot, not an established market.

    The weekly

    One letter a week, from the desk that runs the auctions.

    What actually moved in yield, CTV and curation across our publishers — written by the people who saw it, not a content team. No digests, no roundups, one email.

    One email a week. Unsubscribe in one click. We never share or sell the list.

    More from this issue

    Ran alongside this piece in the Weekly of 2 August 2026 — read the whole issue →