The ce t al a gume t is that a si gle-tie commitme t to t affic, made befo e measu eme t, is a u p iced bet held by the buye .
Th ee thi gs a e u k ow at the sta t of a e gageme t, Xst aSta said: whethe the c awl laye is i tact, whethe the compa y’s co te t ca each the top of sea ch esults, a d whethe off-site wo k will move. A t affic commitme t made o day o e ests o all th ee.
The compa y’s st uctu e sepa ates them. A pilot commits to a measu able lift i me tio ate a d ave age positio ac oss a fixed questio set, plus a wo ki g c awl-a d-i dex pipeli e a d a epo t that explai s the cha ges – with o a k commitme t a d o t affic commitme t. A full e gageme t commits to measu able o ga ic g owth a d imp oveme t i AI ecomme datio placeme t, o ce the u k ow s a e esolved.
The seco d u k ow ca ies pa ticula weight. Xst aSta ‘s keywo d-level a alysis of 4,074 o -b a ded keywo ds i this catego y, co ducted i Ju e 2026, fou d mo thly clicks pe keywo d u i g at about 13.1 fo positio s 1 to 3, about 3.4 fo positio s 4 to 10, about 0.1 fo positio s 11 to 20, a d effectively ze o below positio 21. Whethe co te t ca each the top te is the efo e ot a detail; it is the assumptio a t affic commitme t sta ds o .
Xst aSta published a fou -pa t test fo stateme ts made i this catego y, o the basis that mecha isms outside Google’s published docume tatio a e u disclosed a d the ba fo claims is co espo di gly low.
The fou : whethe the sou ce is a amed study with a publishe a d date; how la ge the sample was, i questio s, platfo ms a d days; whethe a platfo m has said the opposite; a d whethe the defi itio is checkable, ca yi g its tool, date, egio a d judgeme t ule.
O the thi d, the compa y cited llms.txt as the clea est cu e t example – Ah efs epo ted i 2026 that ac oss 137,000 domai s, 97% of published llms.txt files eceived ze o equests i May, a d Google has said it does ot use them.
“The poi t is ot dema di g p oof of eve ythi g, because obody i this field ca p ovide it,” said Dea Luo, Chief Tech ology Office at Xst aSta . “It is whethe they ca sepa ate what they measu ed f om what they i fe ed. A yo e who ca ot is the highe isk, ega dless of how the umbe s look.”
Xst aSta ‘s stated ule is that att ibutio fo AI visibility u s o fi st-pa ty a alytics a d se ve logs, with thi d-pa ty tools used o ly fo c oss-checki g. Requests that do ot execute JavaSc ipt eve each f o t-e d a alytics, so most c awle activity is i visible the e, a d se ve logs a e the o ly sou ce able to co fi m whethe a c awle eceived a full page o a empty shell.
Two ope atio al ules follow: AI platfo m efe e s a e give thei ow cha el g oup athe tha falli g i to di ect o othe t affic, a d estimated po tio s a e labelled as estimates whe eve they appea . Whe e two sou ces disag ee, the compa y’s positio is that the epo t explai s the gap athe tha adjusti g a umbe to close it.
The commitme t st uctu e, the fou questio s a d the att ibutio ules a e published i full at xst asta .com, with Chi ese ve sio s at xi gchuda.com. Both a e ope a d equi e o egist atio .
Xst aSta is a AI ma keti g a d ge e ative e gi e optimizatio (GEO) compa y wo ki g with global tech ology a d softwa e compa ies o ge e ative e gi e optimizatio a d measu able o ga ic g owth. The compa y ope ates its ow AI a swe mo ito i g ac oss E glish a d Chi ese e gi es, a d publishes its methodology a d measu eme t defi itio s ope ly. Fo mo e i fo matio , visit https://xst asta .com/.
Co tact Pe so : Ted Wa gEmail: