Same Story, Different Headline: What Duplicate Coverage Reveals About How Archaeology News Actually Travels

On 23 August 2026, an audit of that week's automated import surfaced two entries that should not have existed: a second listing for the sandstone lintel naming Ramses II "Lord of the City of Masn," and a second listing for the 1,300-year-old industrial complex beneath a Ramat Gan parking lot. Both discoveries were real. Both had already been logged five days earlier, scored, and published. Now they were back, filed under new story IDs, pulled in by a second research pass that had no memory of the first.

A Boundary Day, Counted Twice

The two runs responsible were logged four hours apart on 16 August: one covering a one-day window (15–16 August), the second a same-day catch-up covering zero days (16–16 August). Archaeonews's weekly import always sets a run's start date to the previous run's end date, so consecutive windows meet at a single day — and that day belongs to both. When the first run's window and the catch-up run's window both landed on 16 August, each treated it as unclaimed territory and researched it independently. The Masn story survived as story 1925, sourced from The National, scored 100. The Ramat Gan story survived as story 1926, sourced from Haaretz, scored 81. Their lower-scored twins, pulled in through different URLs by the second pass, were deleted by hand once the overlap was spotted.

Independent Discovery Isn't Really Independent

What makes the duplication instructive is why the second pass found anything at all. Each import run works from a fresh prompt: research archaeology news published in a given window, return verified sources. It carries no memory of what a previous run already found. That would be harmless if two runs never touched the same day — but when they do, the second search is not hallucinating a story. It is doing exactly what it is supposed to do: finding real coverage of a real excavation, published on a real date inside its window. The site's deduplication logic compares titles and URLs, not underlying events, so a differently worded headline from a second search clears the filter even though the discovery underneath is identical.

This is also, not coincidentally, how a lot of archaeological news actually reaches the public. A find surfaces first through the institution closest to it — the Egyptian Ministry of Tourism and Antiquities, the Israel Antiquities Authority — then radiates outward through outlets that each frame it differently, on their own schedule, sometimes days apart. An aggregator built to catch "new" coverage in a given window will, sooner or later, catch the same institutional announcement twice, simply because the announcement itself was never a single event to begin with. It was a press release that several newsrooms independently decided was worth a story.

Fixing the Calendar, Not the Filter

The instinct is to fix this at the point of detection — tighten the title-matching, compare descriptions, cluster more aggressively. That would treat the symptom. The actual defect is upstream, in how the date windows are drawn: making one end of each window exclusive, so a boundary day belongs to exactly one run, removes the double-claim entirely rather than trying to catch it after the fact. That fix has not yet been made; for now, the boundary case is caught by hand, the same way these two duplicates were.

There is something useful in watching an aggregation bug reproduce, in miniature, a feature of the news ecosystem it was built to track. Detecting near-duplicate stories in a stream of institutionally sourced news is a known hard problem, not a bookkeeping detail — and the tendency of press-release-driven science coverage to fragment across outlets that each believe they are reporting something new is exactly the behavior the aggregator's boundary-day bug accidentally reconstructed. The fix belongs in the calendar. The pattern it exposed belongs to how this kind of news has always moved.

References

1. Lewis, J., Williams, A., & Franklin, B. (2008). A Compromised Fourth Estate? UK news journalism, public relations and news sources. Journalism Studies, 9(1), 1–20.
2. Allan, J. (2002). Introduction to Topic Detection and Tracking. In Topic Detection and Tracking (pp. 1–16). Springer.
3. Theobald, M., Siddharth, J., & Paepcke, A. (2008). SpotSigs: Robust and Efficient Near Duplicate Detection in Large Web Collections. Proceedings of SIGIR '08.
4. Davies, N. (2008). Flat Earth News. Chatto & Windus.