You should usee UNITE as your primary database for fungal ITS sequences.

If UNITE isn’t available, use the "Targeted Loci: ITS" option in NCBI BLAST instead of `core_nt`.

For accuracy always cross-check top hits for voucher specimens or cultured strains to ensure reliability.

For fungal ITS, don’t rely on NCBI core_nt (nt) alone. Use a curated ITS database first (e.g., UNITE or NCBI RefSeq Targeted Loci – Fungi ITS) for your primary identification, then cross-check interesting hits against nt for breadth. That gives you the consistency you noticed without the mislabels and duplicates that make nt feel “random.”

Why your nt results looked inconsistent?

• Redundancy & misannotation: nt mixes RefSeq, GenBank, environmental/cloned reads, old names, synonyms, and occasional misIDs. The same sequence can be uploaded under different species names; you’ll see many 100% matches pointing to different taxa.
• Uneven metadata: lots of entries lack voucher links, type status, or verified taxonomy, so top hits can jump around between searches/filters.
• Coverage effects: partial ITS1/ITS2 queries can match different regions in different records, reshuffling “best hit” order.

What to use instead and how...

Primary search (curated):

• UNITE (recommended): purpose-built for fungal ITS with curated taxonomy and Species Hypotheses (SH). Great for consistency and for OTU/ASV work.
• NCBI RefSeq Targeted Loci – Fungi ITS: a stricter, curated subset of ITS; slightly smaller than UNITE but integrated in NCBI tools.

Secondary check (broad coverage):

• NCBI nt (Core Nucleotide): use after you have a candidate ID to see broader synonymy or newly added sequences. Apply filters: “RefSeq only”, “exclude environmental/uncultured”, and if possible “sequences from type material.” This makes nt much more trustworthy.

Practical BLAST tips for ITS:

• Tool: BLASTn (megablast) for close matches; switch to discontiguous BLAST if divergence is high.
• Report both % identity and query coverage. For many fungal groups, species-level calls often need ≥98.5–99% identity with ≥80–90% coverage (group-dependent); genus-level can be ~94–97%. Avoid species calls on short fragments.
• Trim primers/adapters and screen chimeras (e.g., VSEARCH/UCHIME) before BLAST—chimeras can give split hits.
• Know your region: ITS1-only or ITS2-only can bias matches; full ITS (ITS1-5.8S-ITS2) is ideal.
• Check voucher/type links in top hits; prefer records tied to type material or well-curated vouchers.
• If the species call matters (e.g., pathogen), align your query with top curated hits (MAFFT) and do a quick tree (IQ-TREE/RAxML) for context.

Muhammad Ahsan Thank you so much for the answer! I'll defenietly try UNITE for my next stop!