You pull Bing results to see what Copilot might read. Ask for the same query twice, and you can get two different internets.

Kevin Indig, a growth advisor, says 55.7% of the Bing results in his query sample looked poisoned for rank trackers, against 10.1% on Google for the same queries. He gave no sample size, tool or scoring method in the post.

We ran his example query through three collection tools on November 3, 2026. SerpApi returned ten Microsoft account pages, then ten soccer pages, then a clean answer, depending on its settings. My read is that one collected Bing result set is a weak stand-in for what a user or Copilot retrieves.

Indig posted his claim in a short thread on X, and we tested it with his own example.

Indig opened by saying Bing poisons rank tracker results too. Ten minutes later he added that the query was junior sipp, a UK pension product that helps children save for retirement. His read of why it matters: "Maybe trackers are not aware that Bing also poisons results or they're focused more on Google because that's what most people track."

By poisoned, he and the trackers mean a results page with little to do with the query. Google has the same symptom on record. Search Engine Journal reported that SE Ranking found unrelated Google results that mostly followed one word of the query. Almost all began at position 9 or worse. Nozzle put its own figure on page 2 at 12.6% of the desktop searches it could judge. Both firms sell rank tracking software. SE Ranking adds that a site which normally sits in those lower slots can look missing from a tracker report while its rank is unchanged for people who search. SEJ says their rates cannot be compared with each other or with Indig's.

Our check gave three different answers

We ran junior sipp through SerpApi, DataForSEO and a Firecrawl page fetch. SerpApi with cc=gb returned ten Microsoft account pages on the first call. The same setting with caching off returned ten soccer pages, among them Premier League tables, BBC Sport and ESPN. SerpApi with no country set and caching off returned an on-topic page with a Copilot answer. DataForSEO set to the United Kingdom and a Firecrawl fetch from a GB location were both on topic.

Four other queries went through SerpApi Bing and came back on topic: what is a roth ira, best running shoes for flat feet, stocks and shares isa and how to bleed a radiator. A fifth, business bank line of credit, returned three on-topic organic results. So 2 of the 8 SerpApi Bing calls returned unrelated pages, both for junior sipp with cc=gb. That is a handful of calls, not a rate.

The bad pages carried Bing tracking links with the original query inside them, which points to Bing and not to a parser error. If a tracker stored those soccer pages, a relevant page would look absent from that collected result set, whatever its real rank. Disclosure: my agency, ViewEngine, sells AI visibility work and builds its own tracking tools.

The old-news objection

Lily Ray, an SEO, replied on X with a short question: "Wasn't this the case for a while on Bing?" If it was, Indig's rate may not be a recent jump. It also leaves open how long scraped Bing has been unreliable for tracking.

My guess is that tools which read scraped Bing to estimate what Copilot retrieves inherit the same junk. Our test only shows that one query returned three different result sets in one day, depending on the tool, the cc=gb setting and whether the cache was on. Until someone publishes Indig's method, I would treat any single Bing pull as one draw from a noisy source, and collect it twice from different tools before it goes in a report.

Free newsletter

The Sherpa Brief

One email a week on what changed in AI search, what it means, and what to do about it. Written for marketers, SEOs, and publishers.

Free. Unsubscribe in one click.