How Should an Economic Development Organization Prepare for the Screening That Happens Before the RFI?
Michael Etheredge is VP of Product Development at QuestionFuel, where he leads how the firm measures and corrects the way AI answer engines describe companies. He was featured by the Southern Economic Development Council on how AI is changing the way companies research communities and regions.
Site selectors have always screened communities before they send an RFI. A July 2026 Area Development article argues that the screen now runs on public sources and ChatGPT style assistants, and that it can finish in an afternoon. This guide takes that claim at face value as a risk to test, and shows how an economic development organization can prepare.
Key takeaways
- One consultant writes that screening can now finish in an afternoon, before an RFI is drafted, but his article offers no data on how often that happens.
- Early evaluators may review 30 sites, so a site that is hard to evaluate may never receive a serious evaluation.
- Plain tables and short dated summaries are easier to read than brochures and long report bundles, though we have not tested how ChatGPT handles each format.
- Prepare a short, dated brief that states your likeliest elimination reasons and the plan, timeline, and funding behind each.
Quick answer: Prepare for screening before the RFI by publishing one dated fact record in plain text and tables, summarizing each study in a sentence, and stating your three likeliest elimination reasons with a plan, timeline, and funding for each. Then ask ChatGPT about your community three times and mark each fact a match, gap, or error.
What changed about the front end of site selection?
According to one consultant, the screening that used to happen after an RFI now happens before it.
In a Q3 2026 article published July 15, 2026, Devin Hillsdon-Smith of Hyphen Strategies describes the older sequence: a consulting team begins with about two hundred communities, filters by geography, labor market size, infrastructure, and incentives, and arrives at a short list of 25 to 30 communities to contact. He writes that "AI has collapsed that timeline from weeks to hours," and that communities can be eliminated "in an afternoon" before they know a project exists.
The University of North Carolina School of Government describes the same early stage without any mention of answer engines. In its three phase account, Phase 1 screens RFI responses and reduces hundreds of communities to about 10 to 15. The difference is where that screening begins.
Here is what we cannot say. Hillsdon-Smith's article is one practitioner's view, and it gives no count of how many projects work this way. We found no published measurement either. So treat the claim as a risk worth testing, not as an established fact.
What does an evaluator do with the information you send?
An evaluator with limited time looks for facts that are clear, consistent, and easy to verify.
Lindsey Cannon of Quest Site Solutions wrote in Area Development on August 26, 2026 that RFIs often give communities only days, or a week or two, to compile information. She notes that an evaluator doing an initial review may be looking at 30 sites. Her central line is "A site that is difficult to evaluate may never receive a serious evaluation at all."
She describes four common problems:
- •Information overload. Communities send map sets, reports, brochures, and studies as a bundle of attachments with unclear file names.
- •Unreviewed submissions. The contact who submitted the information has not read it. In her example, a utility timeline was longer than the project timeline, and the contact was surprised when the site was eliminated.
- •Inconsistency. The more discrepancies appear across documents, the harder a property is to evaluate.
- •Marketing without facts. Photography and promotion give context but do not prove a site meets criteria.
Her firm sells site certification and economic development services, so read her advice as a practitioner's view. The pattern she describes matches what the Kansas Department of Commerce training deck lists under research and data: accurate data to respond to RFIs, and a website that is updated.
What might a screen find that your own materials do not say?
It may find public records and news you did not write, and they can tell a more complicated story than your marketing.
Hillsdon-Smith lists the kinds of sources an assistant could read: city council minutes, which may record local opposition or zoning disputes, utility rate histories, environmental compliance records, permit timelines, infrastructure bond issuances, and social media sentiment. He gives an example prompt a site selector might use: ask for the top three reasons to eliminate a location for a 500 employee manufacturing facility.
He is careful to add that such profiles are imperfect. Assistants can misattribute data or give recent negative news too much weight, and a good site selector uses the profile as a starting point. But he argues the profile shapes which questions get asked.
For an organization, the useful move is to expect the likeliest elimination reasons and address them in writing first.
AI Narrative Drift: the gap between what is currently true about a company and what answer engines say about it, because those answers update slowly and imperfectly when facts change.
AI Narrative Drift is the name for that gap. When a council minute or an old news item outweighs a current fact, the answer may drift from your documented record.
How do you write a known weaknesses brief?
Write a short page that states each likely elimination reason, the current facts, and the plan behind it.
Danielle Campolo of Strategic Development Group puts the standard this way in a September 23, 2026 Area Development article: if there are constraints, is there a plan, "with a timeline and funding in place," to address them? She adds, "Companies can work around limitations. They can't work around uncertainty."
- •List your three likeliest elimination reasons. Use your own experience and your RFI history. Common ones in the sources include power capacity, utility timelines, permitting speed, and workforce gaps.
- •State the current fact with a date. For example, available capacity today and the date it was confirmed.
- •State the plan, the timeline, and the funding. If funding is not committed, say so.
- •Name who owns each item and who has reviewed it. Cannon's example shows what happens when the contact has not read the material.
- •Add a checked on date and update the page whenever a fact changes.
This brief does not replace a conversation with the utility or the permitting office. It makes sure the written version agrees with what staff will say.
Does the format of your information matter?
Possibly, and the one source that raises it also says the effect is not measured here.
Hillsdon-Smith writes that PDFs, image heavy pages, and information inside interactive platforms are harder for an assistant to parse than plain HTML tables and machine readable text. We have not tested this, and we found no published test of it. It is consistent with Cannon's advice to make facts easy to review, and it costs little to follow.
| Format | Possible advantage | Possible risk | Practical step |
|---|---|---|---|
| Plain HTML table | Facts are readable as text | Needs upkeep when facts change | Put core site and community facts here, with dates |
| Short dated summary | Conclusion is quick to scan | May oversimplify if not careful | Summarize each study in one sentence, with its year |
| Designed brochure or PDF | Good for visits and print | Facts may be harder to extract | Keep the same facts in text on a page |
| Interactive map or database | Useful for browsing | Content may not be readable as text | Mirror key facts in a text page |
Cannon's example summary is "Phase I Environmental Site Assessment completed in 2025 identified no recognized environmental conditions." The goal, she writes, is not to replace the technical documentation but to make its conclusions immediately accessible.
How do you test what ChatGPT says before a project does?
Compare its answers to your fact record, three times each.
- •Write your fact record first. Include sites, power, water, workforce, permitting timelines, and incentives, each with a date.
- •Write five prompts. Include a broad one, such as "What are the reasons to eliminate [community] for a 500 employee manufacturing facility?" and four fact checks, such as "What power capacity is available at [site name]?"
- •Run each prompt three times in new chats. Answers can vary between runs, as explained in Why Does ChatGPT Give Different Site Selection Shortlists to the Same Question?.
- •Mark each fact a match, a gap, or an error. Save the engine and date.
- •Fix the published record first. Then ask partners who control other listings to correct theirs, and repeat the test.
This is the Detect step of the DRIFT Framework. The wider method is in How to Audit What AI Says About Your Region, and the site level version is in How to Describe a Certified Industrial Site So Answer Engines Get It Right.
What is not known yet?
- •We found no published study of how often screening happens before an RFI.
- •We found no published test of whether format changes what ChatGPT says about a community.
- •Hillsdon-Smith's and Cannon's articles are trade publication views from consultants.
- •We have not tested any specific community, and we do not claim to know what ChatGPT says about yours.
When is this not the right approach?
It is not the right first step when the underlying problem is physical, not informational.
Area Development's 2026 survey quotes consultant Gray Swoope saying that markets fall out of consideration early "when utilities cannot provide a clear path to power on the client's timeline." A better description cannot fix a power timeline that is too long. It is also not the right fit if your facts are not yet documented, or if confidentiality limits what you can publish. In those cases, fix readiness and agree what can be public before you write anything.
Frequently asked questions
Is the RFI still important?
Yes. Hillsdon-Smith writes that the RFI still matters, as do site visits and relationships. His point is that the RFI may no longer be where the first elimination happens.
What should we put in plain text on our website?
Put the facts a site selector screens on: sites and buildings, power and water capacity, workforce programs, permitting timelines, and incentives, each with a date. Keep the wording the same everywhere they appear.
Do we need to publish full due diligence reports?
No. Cannon recommends summarizing conclusions, with the study name and year, and sharing the full reports when a prospect asks. What you publish is your decision.
Can we control what ChatGPT says about our community?
No. You can control the accuracy, dates, and consistency of your own record and test the result. We do not claim more than that.
Sources and references
- •Devin Hillsdon-Smith, Hyphen Strategies, Area Development, The RFI Is No Longer the Starting Line, Q3 2026, published July 15, 2026, checked October 23, 2026
- •Lindsey Cannon, Quest Site Solutions, Area Development, Why Great Sites Get Eliminated: The Critical Role of Information Presentation in Site Selection, Q3 2026, published August 26, 2026, checked October 23, 2026
- •Danielle Campolo, Strategic Development Group, Area Development, Ready to Show: Five Things Every EDO Should Be Able to Answer Right Now, Q3 2026, published September 23, 2026, checked October 23, 2026
- •University of North Carolina School of Government, Community and Economic Development, Perspectives on the Business Location and Site Selection Process, November 7, 2025, checked October 23, 2026
- •Area Development Research Desk, 40th Annual Corporate and 22nd Annual Consultant Site Selection Survey Results, Q1 2026, checked October 23, 2026
- •Kansas Department of Commerce, Office of Rural Prosperity, Economic Development 101 training deck, December 2021, checked October 23, 2026
Related reading: How Do Answer Engines Describe Regions? A Guide for Economic Development Organizations · How to Describe a Certified Industrial Site So Answer Engines Get It Right · Does AI Describe Your Region by the Criteria Site Selectors Now Use? · How Does AI Describe Your Region to Companies Deciding Where to Build?