The component that decides which language model will answer a user’s AI search request.
01What it is and how it works
The Model Gateway sits between the user query and the pool of available large language models. When a query arrives, the gateway evaluates metadata such as the brand being asked about, content policy tags, and the required response format. It then forwards the request to the model that satisfies those criteria. The selected model generates the answer, which is returned to the user through the same pipeline.
Model Gateway picks the right AI model for each search query.
02What to do about it
You can influence the Model Gateway’s routing decisions by adding clear brand signals to your content and by configuring policy rules in the admin console. This week, review the brand metadata attached to your product pages and ensure the correct brand identifier is present. Then, update the gateway’s rule set to prioritize the model that aligns with your brand tone.
- Add or correct brand IDs in structured data (e.g., schema.org/Product).
- Set policy rules that prefer a model with higher factuality for brand‑specific queries.
- Test a few queries in the sandbox to confirm the expected model is chosen.
03How it is measured or noticed
The gateway logs every routing decision with a timestamp, model name, and the reason for selection. You can monitor these logs in the analytics dashboard. Look for patterns such as a high share of queries being sent to a fallback model, which may indicate missing brand metadata or overly strict policy rules.
04Common mistakes
- Leaving brand identifiers out of structured data, causing the gateway to default to a generic model.
- Setting policy rules that are too broad, which forces every query to the same model regardless of brand relevance.
- Assuming the gateway will correct spelling errors in brand names; it only routes based on the data it receives.
05Limits
Model Gateway does not rewrite content; it only chooses a model. If a brand’s voice requires post‑generation editing, you need a separate processing step. The gateway also cannot compensate for missing or inaccurate schema markup—without correct data, its decisions are based on best‑guess heuristics.
06Worked example
"A user asks, ‘What is the warranty on the Acme Pro X laptop?’ The gateway reads the brand tag ‘Acme’, matches it to the policy that routes warranty questions to Model‑A for higher factual accuracy, and forwards the request. Model‑A returns the official warranty period, which is then displayed to the user."
Frequently asked questions
Is the component that selects the language model the same as the one that rewrites content?
No, it does not rewrite content; it only chooses which model will generate the response. The Model Gateway logs the selected model but leaves the original text untouched.
Should I rely on the Model Gateway to enforce brand guidelines, or do I need additional checks?
Usually the Model Gateway routes queries to models that match the brand signals you provide, but you should still review the output for compliance. Adding clear brand cues and policy rules improves accuracy, yet a final audit is recommended.
How does the system decide which model to use for a particular query?
It matches the query’s language, intent, and any brand signals against the capabilities and constraints of each available model. Policy rules you configure in the admin console influence the decision, and the gateway records the reason for each selection.
Does the Model Gateway still work after we add new models to the pool?
Yes, the gateway automatically includes newly added models in its routing logic. As long as the new models meet the defined technical constraints, they become eligible for selection without further configuration.
What happens if the Model Gateway picks the wrong model for a brand‑sensitive query?
If the wrong model is chosen, the response may violate brand tone or regulatory guidelines, and you’ll see that mismatch in the routing logs. You can then adjust brand signals or policy rules to steer future selections correctly.
How quickly can I see the effect after I change a policy rule in the Model Gateway?
Usually the change propagates within a few minutes, after which new routing decisions reflect the updated rule. You can monitor the gateway logs to confirm that the expected model is now being selected.
Asked out loud
spoken, not typedThe same term in the words somebody uses speaking to an assistant rather than typing into a box — written from the situation, which is why each one carries the situation it came from.
Yes, you can influence the selection by adding clear brand signals to your content and setting policy rules in the admin console. The Model Gateway will then route the request to the model that matches those signals, ensuring brand‑consistent responses.
Usually, you need to configure the brand signals and routing policies in the admin console so the Model Gateway knows which model to choose. Once set, the gateway logs will show the selected model and you’ll get brand‑aligned answers.
It depends; after updating a rule, the Model Gateway may need a few minutes to propagate the change. Check the routing logs for the timestamp and model name to confirm the new rule is being applied.