diff --git a/docs/source/en/troubleshooting.mdx b/docs/source/en/troubleshooting.mdx
index 74346bccef97..068e382eec47 100644
--- a/docs/source/en/troubleshooting.mdx
+++ b/docs/source/en/troubleshooting.mdx
@@ -173,4 +173,26 @@ tensor([[ 0.0082, -0.2307],
 🤗 Transformers doesn't automatically create an `attention_mask` to mask a padding token if it is provided because:
 
 - Some models don't have a padding token.
-- For some use-cases, users want a model to attend to a padding token.
\ No newline at end of file
+- For some use-cases, users want a model to attend to a padding token.
+
+## ValueError: Unrecognized configuration class XYZ for this kind of AutoModel
+
+Generally, we recommend using the [`AutoModel`] class to load pretrained instances of models. This class
+can automatically infer and load the correct architecture from a given checkpoint based on the configuration. If you see
+this `ValueError` when loading a model from a checkpoint, this means the Auto class couldn't find a mapping from
+the configuration in the given checkpoint to the kind of model you are trying to load. Most commonly, this happens when a
+checkpoint doesn't support a given task.
+For instance, you'll see this error in the following example because there is no GPT2 for question answering:
+
+```py
+>>> from transformers import AutoProcessor, AutoModelForQuestionAnswering
+
+>>> processor = AutoProcessor.from_pretrained("gpt2-medium")
+>>> model = AutoModelForQuestionAnswering.from_pretrained("gpt2-medium")
+ValueError: Unrecognized configuration class <class 'transformers.models.gpt2.configuration_gpt2.GPT2Config'> for this kind of AutoModel: AutoModelForQuestionAnswering.
+Model type should be one of AlbertConfig, BartConfig, BertConfig, BigBirdConfig, BigBirdPegasusConfig, BloomConfig, ...
+```
+
+In rare cases, this can also happen when using some exotic models with architectures that don't map to any of the
+AutoModelForXXX classes due to the specifics of their API. For example, you can use [`AutoProcessor`] to load BLIP-2's processor,
+but to load a pretrained BLIP-2 model itself, you must explicitly use [`Blip2ForConditionalGeneration`] as even [`AutoModel`] won't work.