Sixth on the broad query. First on the one it could own.

Nothing technical. Both queries land on the same site, with the same crawler access, the same structured data and the same page structure. The difference is that fxstrength.org includes gold, silver and Bitcoin in its strength meter, and most currency strength meters don't. On the broad query it's one of many interchangeable tools. On the narrow one it's the obvious answer.

That's the pattern worth looking for in any business: most are competing on a term they can't win, while sitting on one they could own and have never written a page for.

Google's first page has ten places. An AI answer usually names two or three. The old SEO plan, climbing from ninth to fifth, doesn't carry over: fifth place in an AI answer means not being named at all. The only positions that exist are the ones that get named, and the easiest way to be named is to be the clearest answer to a specific question.

On its own, yes, which is why the test isn't “where can you be first?” but “where can you be first and is someone ready to buy asking?” The gold queries are small, but 19 of 30 searchers clicked. They're people looking for exactly what the site offers.

And the narrow query is a beachhead, not a destination. Winning where you can win builds the authority, the consistent description and the citations that eventually let a site compete on the term it couldn't touch at the start.

92% of the site's Google clicks come from branded searches, people typing the site's name. Google works as a way back for visitors who already know it, not as the place new visitors discover it. New discovery is coming from elsewhere, increasingly from AI assistants: between 16 July and 10 October 2026, ChatGPT sent the site 1,487 visitors, against 739 from Google and Bing organic search combined.

What it means for you: if you judge your visibility by Google rankings and organic traffic, you may be measuring the wrong channel.

On Bing, the site's main head term had 7,486 impressions at an average position of 6.2, a first-page result, and a click-through rate of just 0.67%. The likeliest explanation is that the question is answered on the results page itself, by Copilot, partly using this site's own content. The visitor gets the answer and has no reason to click.

What it means for you: falling click-through on a page that still ranks isn't necessarily failure. The traffic may have turned into a citation, and you'll only see that if you look at citation data, not just clicks.

In the 75 days to 7 October 2026, Bing logged 19,901 AI citations of the site, about 265 a day. Over a near-identical window (23 July to 6 October), Google showed it in AI Overviews 130 times. The two figures aren't the same measure, citations against impressions, so they shouldn't be divided into a neat ratio. But the gap is too large to be a measurement quirk.

The asymmetry is a finding in itself: Bing Webmaster Tools reports how often AI answers cite your site, while Google Search Console reports only impressions and clicks for its AI features, with no citation count. Only one of the two tells publishers when AI quotes them, which is part of why so few have looked.

ChatGPT's web search has relied heavily on Bing's index, and Bing powers Copilot. Yet almost no one optimises for Bing. Bing Webmaster Tools is free, reports AI citation data directly, and most companies have never opened it.

What it means for you: submit your sitemap to Bing Webmaster Tools and check its AI performance report. It's often the cheapest, most overlooked step in this whole field.