When buyers ask AI who to hire, are you in the answer?

There is no AI ranking. There are several engines that disagree.

ChatGPT runs its own index alongside Bing, Google and vendor feeds. Claude searches through Brave. Perplexity runs a separate index of its own. Google's AI features run on the ordinary Google index. Each one retrieves differently, trusts different sources, and answers the same question with different companies in it.

So "we will get you into AI" is not a service anyone can sell honestly. Being in one engine tells you very little about the others.

We hold ourselves to the same standard on what we do not know. Grok and Meta AI matter in India and we test them, but xAI publishes no crawler documentation at all, so anything we told you about controlling it would be inference dressed as fact. It is not in the table below for that reason.

Do not buy AI citations as a traffic channel. That is the honest pitch.

Cloudflare measured how many pages each platform crawls for every visitor it sends back. In the week of 19 to 26 June 2025 the spread ran from Anthropic at roughly 70,900 pages crawled per referral down to far lower ratios elsewhere. Cloudflare notes that native app traffic carries no referrer header, so the true ratios are likely better than measured, by an unknown amount.

Ahrefs analysed 4 million AI Overview URLs across 863,000 result pages in March 2026 and found 37.9% of cited URLs also appeared in the first ten blocks of the ordinary page. Blocks are not organic positions: ads, featured snippets, People Also Ask and video packs each count as one, so the window is tighter than the top ten results.

We previously presented that beside an earlier Ahrefs figure of 76% as a fall over time. That comparison does not hold. The earlier study looked at the top three citations per answer against organic rank; the later one looked at every cited URL against blocks. Different scope, different denominator, so the two numbers describe different measurements rather than a trend. Ahrefs sells SEO tooling, and both caveats above are ours.

Read together: ranking well and being cited overlap only partly, and nothing here supports treating citations as a dependable source of visits. Crawl-to-referral ratios do not measure visits per citation, so they cannot say how much traffic a single citation sends. The value is being the name a buyer hears when they ask, which shows up later as someone typing your business into a search box. That is what we measure.

One screenshot is not evidence. A rate is.

We fix a set of prompts your buyers would really type, run them across the engines your buyers really use, in fresh sessions, and record four separate things: whether you can be retrieved at all, whether you appear in that engine's results, whether you are linked in the answer, and whether you are named in the text. Where we have run this, the gap has most often been between the second and third. That is an observation from a small number of checks, not a measured rate, and it usually points to a passage problem rather than an authority problem.

We also record what got cited instead of you, because that list is the plan. Results are reported as rates across runs, never as a single answer, and we re-run the same set monthly so the comparison means something.