We have analysed all the good quality XMM–Newton data publicly available for the bright ultraluminous X-ray sources Holmberg IX X-1 and NGC 1313 X-1, with the aim of searching for discrete emission or absorption features in the Fe K band that could provide observational evidence for the massive outflows predicted if these sources are accreting at substantially super-Eddington rates. We do not find statistically compelling evidence for any atomic lines, and the limits that are obtained have interesting consequences. Any features in the immediate Fe K energy band (6–7 keV) must have equivalent widths weaker than ∼30 eV for Holmberg IX X-1 and weaker than ∼50 eV for NGC 1313 X-1 (at 99 per cent confidence). In comparison to the sub-Eddington outflows observed in GRS 1915+105, which imprint iron absorption features with equivalent widths of ∼30 eV, the limits obtained here appear quite stringent, particularly when Holmberg IX X-1 and NGC 1313 X-1 must be expelling at least 5–10 times as much material if they host black holes of similar masses. The difficulty in reconciling these observational limits with the presence of strong line-of-sight outflows suggests that either these sources are not launching such outflows or they must be directed away from our viewing angle.