I asked this question when looking for background information for one of my PhD projects and I gained a lot of insight since, which I wanted to share. If you are interested, the project outcome was published recently$^1$. In short - autofluorescence of filter paper originates both from OBAs as well as from cellulose fibrils itself.
There is several sources of fluorescence in papers and other lignocellulosic materials including both optical brightening additives (OBA) as @Klaus pointed out as well as cellulose itself. In case of biomass, lignin is another strongly fluorescent component, that is however removed in manufacturing process of paper.
There exist few earlier literature accounts proving that pure cellulose itself exhibits weak near UV fluorescence, including this paper by Olmstead and Gray$^2$. Here Emission spectra of various cellulose samples at $\lambda_{ex}=320$ nm was recorded (Fig. 1). Bacterial cellulose is know to be fully lignin-free, so there is no doubt that the spectra comes from cellulose fibrils themselves.

The absorbance spectrum of multi-component biomass materials is more complicated. I measured the Raman spectra of filter paper and two unbleached biomass samples and confirmed the cellulose was present in all of them (Fig. 2C). Lignin was absent in filter paper, but present in biomass samples. Next, I performed excitation/emission autofluorescence spectral scans of same samples in UV-VIS range (Fig. 2D). I found dominant peak at 360/420 $\lambda_{ex} / \lambda_{em}$ present in all specimen that can only be attributed to cellulose. Filter paper had a a few more sharp fluorescence peaks at lower wavelengths, absent in unbleached biomass, that I attributed to OBAs. Biomass sampled exhibited a broad peak shoulder at higher wavelengths that originates from lignin and was absent in delignified filter paper.
Fig. 1 Pretreated biomass substrate characterization. Side and top view of cylindrical discs (6 mm in diameter) produced from (A) napier and (B) miscanthus perennial grass. (C) Raman spectra of pretreated biomass substrates using 568 nm excitation. Bands at 380, 435, 1095, 1123, 1377, and 2985 cm$^{−1}$ were attributed to cellulose, with lignin contribution visible at 1600 cm$^{−1}$. Spectra were background-corrected and vertically offset for clarity. Excitation/emission autofluorescence spectral scans of (D) filter paper, (E) miscanthus, and (F) napier grass exhibited a prominent cellulose peak at ≈365/430 nm $\lambda_{ex} / \lambda_{em}$. A lignin shoulder at longer wavelengths was present in miscanthus and napier grass samples.
One question I was unable to answer is what is the molecular origin of cellulose fluorescence ? The origin of lignin fluorescence is well understood $^{3,4}$, with distinct fluorophores formed by aromatic moieties in lignocellulose structure. Cellulose, on the other hand, does not have an aromatic character. I came across a hypothesis that fluorophores are formed when glucose rings in cellulose chains interact with each other via van der Waals and stacking interactions forming cellulose nanocrystals. That would suggest that purely amorphous cellulose should not be fluorescent. I was, however, unable to find studies on that topic.
- Malinowska K, Rind T, Verdorfer T, Gaub HE, Nash MA (2015) Quantifying Synergy, Thermostability, and Targeting of Cellulolytic Enzymes and Cellulosomes with Polymerization-Based Amplification. Anal Chem 87:7133–7140.
- Olmstead JA, Gray DG (1993) Fluorescence emission from mechanical pulp sheets. J Photochem Photobiol A 73:59–65.
- Albinsson B, Li S, Lundquist K, Stomberg R (1999) The origin of lignin fluorescence. Journal of Molecular Structure.
- Radotić K et al. (2006) Component analysis of the fluorescence spectra of a lignin model compound. J Photochem Photobiol B 83:1–10.