I'd spent months optimizing for one search engine. Not on purpose — Google is just where I looked. Search Console was wired into my morning reports, the numbers showed up in Telegram every day, and that was "search performance." Bing Webmaster Tools existed in a browser tab I never opened.
When I finally opened it, the data didn't agree with Google. It actively contradicted it.
On the site, two pages cover the same subject. One is a dated long-form article — the "everything about this topic, updated for 2026" piece. The other is an interactive map that renders the same underlying data geographically: pick a place, see its status. They target the same cluster of queries.
On Google, the article wins decisively — roughly five times the impressions of the map, and a better average position. I'd internalized that as settled: Google had picked the article as the answer, end of story.
On Bing, the map wins. The article is barely present. The map page alone accounts for 56% of every impression the site gets on Bing. The article that dominates Google sits near the bottom of the list.
Same two pages. Opposite order. Two engines looking at identical content and disagreeing about which one deserves to rank.
| Page | Bing | |
|---|---|---|
| Long-form article | Wins | Loses |
| Interactive map | Loses | Wins |
Why an engine "picks" a page
This isn't randomness, and it isn't a quirk of one crawl. The two engines reward different things for the same informational query.
Google rewards the article. Its helpful-content systems lean hard toward substantive, dated, written pages for informational intent. A long article with depth, a clear publish date, and a "2026" freshness signal reads to Google as the authoritative answer to "tell me about this topic." The map page — mostly an interactive widget with thin surrounding text — reads as a tool, not an answer. Google buries it.
Bing rewards the tool. Bing is more literal. When the query itself implies a lookup — anything phrased as a map or a by-location question — Bing matches the page whose entire purpose is that lookup. The map page's title, structure, and intent line up with the query word-for-word, so Bing puts it first. The article, to Bing, is just a wall of text adjacent to the thing the user actually asked for.
Neither is wrong. They're optimizing for different definitions of "best result," and my two pages happened to be perfect representatives of each definition. So each engine crowned a different king.
The inversion showed up at the sub-page level too, even more starkly: one regional reference page ranks on the first page of Bing and sits at position 69 on Google — effectively invisible. Same page. The gap there isn't content; it's authority. Those head terms on Google are owned by encyclopedias, government sites, and AI summaries, and a young page doesn't crack them. Bing's bar is lower, so the same page surfaces. Watching only Google, I'd have called that page a failure. It isn't — it's just winning on the engine I never checked.
The zero-click trap
Here's the part that nearly fooled me twice.
The map page's top Bing queries — the high-volume, head-of-cluster ones — rank around positions 6 to 9 and convert almost no clicks. My first read was "impressions without clicks, so this is worthless ranking."
That might be the wrong conclusion. One likely explanation — and I want to be careful to call it a hypothesis, not a proven cause — is that some of those impressions are being answered inline by an AI assistant: the model pulls the page into its grounding set, writes the answer, and the user never clicks because they already got what they came for. I've written before about why AI citation is becoming the metric that replaces the click, and a high-impression, zero-click, mid-position query is exactly the shape that pattern leaves behind.
Here's the honest limit, though: I can't confirm it from the data I have. The search engine's webmaster API reports impressions, clicks, and positions — it does not expose whether an AI surface cited the page. So "these zero-click impressions are AI citations" is an inference from the shape of the numbers, not something I measured. It's plausible, it fits, and it's the explanation I'd bet on — but it's a bet, and treating it as confirmed would be the exact confidence gap I keep warning about. What I can say for certain is narrower and still useful: zero clicks at position 6–9 does not mean zero value, and reading it as failure would be a mistake.
So "zero clicks" on those queries isn't dead weight. It's the citation surface. The clicks I do get come from longer-tail queries lower in the cluster, ranking higher, where no AI summary intercepts them. Two completely different value mechanisms inside the same page, and you can only tell them apart if you stop treating clicks as the only outcome that counts. (If the GEO/AEO/citation vocabulary is new, I untangled the three terms here.)
What watching one engine hid
I would not have seen any of this from Google alone. Not the inversion, not the page that's invisible on one engine and page-one on the other, not the citation pattern. Google's data is internally consistent and tells a clean story — the article is the answer — and that story is true on Google and false everywhere else.
That's the part that actually unsettled me, and it's bigger than Bing. For years my working definition was Google = search. One dashboard, one set of numbers, one verdict on whether a page worked. But search has quietly become plural — Google, Bing, AI Overviews, Perplexity, ChatGPT, Gemini — and they don't agree, because they don't rank the same way or even reward the same kind of page. My mental model wasn't wrong so much as incomplete: I was reading one instrument and calling it the weather. The inversion was just the first thing that didn't fit, because it was the first time I looked at a second instrument.
The fix was boring and is the actual point: pipe the second engine into the same place as the first. Bing exposes a Webmaster API with the same kind of search-performance data Google's does — totals, top pages, top queries, positions — behind a single API key you generate in the tool's settings. I wired Google Search Console into a daily Telegram briefing months ago; adding Bing was the same shape of work pointed at a different endpoint. Now both engines land in the same thread, weekly, side by side — a short "this week vs last week" and a six-month trend so I can watch the curve. (That curve, for the record, has compounded every month from a near-zero start — visibility I had no idea was accruing because I wasn't looking.)
One gotcha worth saving someone an hour: the related IndexNow endpoint — the protocol that lets you ping Bing and others the instant a URL changes — returned 403 Forbidden for every request until I added a User-Agent header. curl and the browser don't care; the API does. No error message explains it. If your IndexNow submissions are silently rejected, send a User-Agent.
The takeaway
"Rank" isn't one number. The same page can be a winner and a loser at the same moment depending on who's asking, and the two biggest engines can disagree completely about content they're both looking at.
The practical move once you see the split is not to force the pages to fight — I'm not stripping the article of the terms it wins on Google to help the map, or vice versa. They've each won an engine. The right move is to let them, cross-link them so they reinforce instead of compete, and keep both engines on the dashboard so the next divergence isn't invisible for months.
But the durable lesson isn't about Bing at all. It's that treating one engine as the whole of search means your picture is incomplete by definition — and you won't know what you're missing until you instrument the engines you've been ignoring. Mine had a story in it the whole time. I just wasn't looking.
Originally published on tedagentic.com.
Top comments (0)